@windyroad/itil 2.4.2 → 2.4.3-preview.1248

package/.claude-plugin/plugin.json

@@ -497,5 +497,5 @@
     }
   },
   "name": "wr-itil",
-  "version": "2.4.2"
+  "version": "2.4.3"
 }

package/.codex-plugin/plugin.json

@@ -1,6 +1,6 @@
 {
   "name": "wr-itil",
-  "version": "2.4.2",
+  "version": "2.4.3",
   "description": "ITIL problem-management workflows for AI coding agents",
   "author": {
     "name": "Windy Road Technology",

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@windyroad/itil",
-  "version": "2.4.2",
+  "version": "2.4.3-preview.1248",
   "description": "ITIL-aligned IT service management for Claude Code and Codex",
   "bin": {
     "windyroad-itil": "./bin/install.mjs"

package/skills/close-incident/SKILL.md

@@ -134,7 +134,7 @@ Otherwise, after the commit in step 7 lands, drain the release queue so the clos
    - **Primary**: delegate to subagent type `wr-risk-scorer:pipeline` via the Agent tool.
    - **Fallback**: if that subagent type is not available, invoke skill `/wr-risk-scorer:assess-release` via the Skill tool.
 2. Read the returned `RISK_SCORES: commit=X push=Y release=Z` line.
-3. **Drain condition**: if `push` and `release` are both within appetite (≤ 4/25, "Low" band per `RISK-POLICY.md`), AND `.changeset/` is non-empty, proceed to the drain action. Otherwise, skip the drain and report the unreleased state.
+3. **Drain condition**: if `push` and `release` are both within appetite (at or below the effective appetite used by the scorer and gate), AND `.changeset/` is non-empty, proceed to the drain action. Otherwise, skip the drain and report the unreleased state.
 
 **Drain action (non-interactive, policy-authorised per the "Structured User Interaction for Governance-Skill Decisions" architecture rule Rule 6):**
 

package/skills/link-incident/SKILL.md

@@ -117,7 +117,7 @@ Otherwise, after the commit in step 7 lands, drain the release queue so the link
    - **Primary**: delegate to subagent type `wr-risk-scorer:pipeline` via the Agent tool.
    - **Fallback**: if that subagent type is not available, invoke skill `/wr-risk-scorer:assess-release` via the Skill tool.
 2. Read the returned `RISK_SCORES: commit=X push=Y release=Z` line.
-3. **Drain condition**: if `push` and `release` are both within appetite (≤ 4/25, "Low" band per `RISK-POLICY.md`), AND `.changeset/` is non-empty, proceed to the drain action. Otherwise, skip the drain and report the unreleased state.
+3. **Drain condition**: if `push` and `release` are both within appetite (at or below the effective appetite used by the scorer and gate), AND `.changeset/` is non-empty, proceed to the drain action. Otherwise, skip the drain and report the unreleased state.
 
 **Drain action (non-interactive, policy-authorised per the "Structured User Interaction for Governance-Skill Decisions" architecture rule Rule 6):**
 

package/skills/manage-incident/SKILL.md

@@ -307,19 +307,19 @@ Commit the completed work per the "Governance Skills Commit Their Own Completed
    - Incident closed: `docs(incidents): close I<NNN>`
 4. If commit risk is above appetite: this is **framework-mediated, NOT a category-3 one-time-override ask** (the "Skills, agents, and hooks override RISK-POLICY appetite instead of applying it" problem/the ": Apply RISK-POLICY appetite faithfully across all surfaces" release design amendment 2026-06-24 — there is no incident carve-out). Per **the "Auto-apply scorer remediations to reach within appetite — open action-class vocabulary" architecture rule Rule 1b (incident-context scoring)**: an active incident is a risk being realised (Likelihood already 5), so the incident-response change is scored against that live realised-risk baseline — weighing P(the change increases impact) vs P(it reduces impact & likelihood / restores service) + P(it introduces a new incident). If net risk-reducing, it takes the risk-reducing path (the `reducing` bypass — `RISK_BYPASS: reducing`) and proceeds, no ask. This is *better* for restore-service-fast (the "Restore Service Fast with an Audit Trail" user outcome): a genuine hotfix clears via scoring with no consent gate mid-outage. If the change is NOT net-reducing, auto-remediate per the "Auto-apply scorer remediations to reach within appetite — open action-class vocabulary" architecture rule Rule 1 or halt per Rule 5. **MUST NOT commit above appetite; MUST NOT `AskUserQuestion` "commit anyway".** If `AskUserQuestion` is unavailable, the "Structured User Interaction for Governance-Skill Decisions" architecture rule Rule 6 fail-safe applies — skip the commit and report the uncommitted state.
 
-### 15. Auto-release when changesets are queued (the "Governance skills auto-release when changesets are queued" architecture rule)
+### 15. Auto-release when changesets are queued (the "Reuse valid cumulative risk assessments across pipeline actions" architecture rule)
 
 **Skip this step if the skill is running inside an AFK orchestrator.** Orchestrators handle release cadence themselves per the "Inter-iteration release cadence for AFK loops" architecture rule (Step 6.5). When in doubt, defer to the orchestrator by skipping this step.
 
 Otherwise, after the commit in step 14 lands, drain the release queue so the fix actually lands on npm without requiring manual user action.
 
-**Mechanism — delegate, do not re-implement scoring (per the "On-Demand Assessment Skills for Governance Plugins" architecture rule):**
+**Mechanism — reuse a valid cumulative assessment; delegate when needed (the "Reuse valid cumulative risk assessments across pipeline actions" architecture rule, the "On-Demand Assessment Skills for Governance Plugins" architecture rule):**
 
-1. Invoke the release scorer. Two paths are valid:
+1. Use the Step 14 pipeline assessment's separate `push` and `release` scores if it still covers this exact checkout, assessed content, unpushed and unreleased scope, and current effective appetite, and has not expired or drifted. A commit of identical assessed content does not require another scorer invocation. If any condition fails or cannot be established, invoke the scorer again before pushing or releasing. Two delegation paths are valid:
    - **Primary**: delegate to subagent type `wr-risk-scorer:pipeline` via the Agent tool.
    - **Fallback**: if that subagent type is not available, invoke skill `/wr-risk-scorer:assess-release` via the Skill tool. The skill wraps the same pipeline subagent.
-2. Read the returned `RISK_SCORES: commit=X push=Y release=Z` line.
-3. **Drain condition**: if `push` and `release` are both within appetite (≤ 4/25, "Low" band per `RISK-POLICY.md`), AND `.changeset/` is non-empty, proceed to the drain action. Otherwise, skip the drain and report the unreleased state.
+2. Read `RISK_SCORES: commit=X push=Y release=Z` from the valid assessment. Never substitute the commit score for a push or release score. The action-time `push:watch` and `release:watch` gates still check the relevant score, checkout, state, expiry, appetite, and CI; if a gate rejects stale or changed evidence, resolve the checkout binding or rescore before retrying.
+3. **Drain condition**: if `push` and `release` are both within appetite (at or below the effective appetite used by the scorer and gate), AND `.changeset/` is non-empty, proceed to the drain action. Otherwise, skip the drain and report the unreleased state.
 
 **Drain action (non-interactive, policy-authorised per the "Structured User Interaction for Governance-Skill Decisions" architecture rule Rule 6):**
 
@@ -329,7 +329,7 @@ Otherwise, after the commit in step 14 lands, drain the release queue so the fix
 
 **Failure handling**: If `release:watch` fails (CI failure, publish failure), stop and report the failure clearly. Do not retry non-interactively — the user must intervene.
 
-**Above-appetite branch (per the "Auto-apply scorer remediations to reach within appetite — open action-class vocabulary" architecture rule)**: If push or release risk is above appetite (≥ 5/25), the skill MUST auto-apply scorer remediations incrementally until residual risk converges within appetite, OR halt the skill per the "Auto-apply scorer remediations to reach within appetite — open action-class vocabulary" architecture rule Rule 5 if the scorer cannot produce a convergent plan. **The skill MUST NOT release above appetite under any circumstance.** The skill MUST NOT call `AskUserQuestion` as a shortcut out of the auto-apply loop.
+**Above-appetite branch (per the "Auto-apply scorer remediations to reach within appetite — open action-class vocabulary" architecture rule)**: If push or release risk is above appetite (above the effective appetite used by the scorer and gate), the skill MUST auto-apply scorer remediations incrementally until residual risk converges within appetite, OR halt the skill per the "Auto-apply scorer remediations to reach within appetite — open action-class vocabulary" architecture rule Rule 5 if the scorer cannot produce a convergent plan. **The skill MUST NOT release above appetite under any circumstance.** The skill MUST NOT call `AskUserQuestion` as a shortcut out of the auto-apply loop.
 
 **Auto-apply mechanism (the "Auto-apply scorer remediations to reach within appetite — open action-class vocabulary" architecture rule Rule 2):**
 
@@ -355,7 +355,7 @@ Otherwise, after the commit in step 14 lands, drain the release queue so the fix
 - **the "Add `manage-incident` Skill to `wr-itil` Plugin" architecture rule** (`docs/decisions/011-manage-incident-skill.proposed.md`) — incident lifecycle; evidence-first workflow; reversible-mitigation preference; Sev 4-5 lightweight path. Step 6's evidence-gate refactor (2026-04-28) extends the "Add `manage-incident` Skill to `wr-itil` Plugin" architecture rule's evidence-first rule with the documented `Record anyway` audit-trail bypass that mitigate-incident already used (cool-headed-commitment consistency across the two incident skills).
 - **the "Governance Skills Commit Their Own Completed Work" architecture rule** — governance skills commit their own work. Step 14 unchanged.
 - **the "On-Demand Assessment Skills for Governance Plugins" architecture rule** — release scorer delegation pattern. Step 15 unchanged.
-- **the "Inter-iteration release cadence for AFK loops" architecture rule** + **the "Governance skills auto-release when changesets are queued" architecture rule** — release cadence. Step 15 unchanged.
+- **the "Inter-iteration release cadence for AFK loops" architecture rule** + **the "Reuse valid cumulative risk assessments across pipeline actions" architecture rule** — AFK release cadence and attended reuse of valid cumulative risk scores in Step 15.
 - **the "Agent output grounding — no fabricated quantitative estimates, cite + persist + uncertainty" architecture rule** — cost-source grounding. Step 6's audit-trail bypass note preserves grounding by capturing the user's justification at deviation time.
 - **the "Auto-apply scorer remediations to reach within appetite — open action-class vocabulary" architecture rule** — auto-apply scorer remediations. Step 15 above-appetite branch unchanged.
 - **the "Problem 071: Argument-based skill subcommands are not discoverable in Claude Code autocomplete" problem** — skill-split origin (slice 6 — manage-incident is the host with thin-router forwarders for `list`, `mitigate`, `restored`, `close`, `link`).

package/skills/manage-problem/SKILL.md

@@ -1147,23 +1147,23 @@ Commit the completed work per the "Governance Skills Commit Their Own Completed
    - Problem closed: `docs(problems): close P<NNN> <title>`
    - Review/re-rank: `docs(problems): review — re-rank priorities`
    - Fix implemented: `fix(<scope>): <description> (closes P<NNN>)` — include problem file changes (rename to `.verifying.md` + `## Fix Released` section) in the same commit per the "Problem lifecycle — add a Verification Pending status between Known Error and Closed" architecture rule
-5. If commit risk is above appetite: auto-apply scorer remediations per **the "Auto-apply scorer remediations to reach within appetite — open action-class vocabulary" architecture rule Rule 1** incrementally until residual commit risk is within appetite (≤ 4/25), OR halt per the "Auto-apply scorer remediations to reach within appetite — open action-class vocabulary" architecture rule Rule 5 if the scorer cannot converge. **The skill MUST NOT commit above appetite, and MUST NOT call `AskUserQuestion` to ask whether to commit anyway** (the "Skills, agents, and hooks override RISK-POLICY appetite instead of applying it" problem/the ": Apply RISK-POLICY appetite faithfully across all surfaces" release design amendment 2026-06-24 — above-appetite is framework-mediated, never a category-3 one-time-override; same invariant the push/release branch at Step 12 already enforces). The "Structured User Interaction for Governance-Skill Decisions" architecture rule Rule 6 fail-safe (no `AskUserQuestion` available / non-interactive → skip the commit and report the uncommitted state) remains the terminal fallback. This applies only to the risk-above-appetite branch, not to the delegation-unavailable case above.
+5. If commit risk is above appetite: auto-apply scorer remediations per **the "Auto-apply scorer remediations to reach within appetite — open action-class vocabulary" architecture rule Rule 1** incrementally until residual commit risk is within appetite (at or below the effective appetite used by the scorer and gate), OR halt per the "Auto-apply scorer remediations to reach within appetite — open action-class vocabulary" architecture rule Rule 5 if the scorer cannot converge. **The skill MUST NOT commit above appetite, and MUST NOT call `AskUserQuestion` to ask whether to commit anyway** (the "Skills, agents, and hooks override RISK-POLICY appetite instead of applying it" problem/the ": Apply RISK-POLICY appetite faithfully across all surfaces" release design amendment 2026-06-24 — above-appetite is framework-mediated, never a category-3 one-time-override; same invariant the push/release branch at Step 12 already enforces). The "Structured User Interaction for Governance-Skill Decisions" architecture rule Rule 6 fail-safe (no `AskUserQuestion` available / non-interactive → skip the commit and report the uncommitted state) remains the terminal fallback. This applies only to the risk-above-appetite branch, not to the delegation-unavailable case above.
 
 **Multi-commit slice changeset discipline (the "AFK iter `packages/<plugin>/` commits without changesets — orchestrator-main-turn back-fill is fragile recovery, hook-level enforcement preferable" problem Phase 2)**: when a single logical fix lands across multiple the "Governance Skills Commit Their Own Completed Work" architecture rule-grain commits targeting the same plugin (e.g. helper extraction in commit 1, callers wired in commit 2, SKILL note + transition in commit 3 — all `packages/<plugin>/`), author ONE changeset on the first commit in the slice. Subsequent same-plugin commits do NOT need their own changeset — the `itil-changeset-discipline.sh` hook's Check 2b recognises any `.changeset/*.md` already in the unpushed slice scope (`origin/<base>..HEAD` + untracked + modified-not-staged) that targets `"@windyroad/<plugin>": <any-bump>` and allows. This eliminates the per-commit changeset ceremony that previously produced N redundant `.changeset/*.md` files for one logical release entry (changesets-action collapses bump-class at version-package time, so per-commit changesets rendered N near-identical CHANGELOG bullets for one release). Once a changeset hits `origin/<base>` (drained at release time), it no longer counts — a fresh changeset is required for the next slice. Cross-plugin coverage is NOT permitted: an `@windyroad/itil` changeset does not satisfy a `packages/voice-tone/` commit.
 
-### 12. Auto-release when changesets are queued (the "Governance skills auto-release when changesets are queued" architecture rule)
+### 12. Auto-release when changesets are queued (the "Reuse valid cumulative risk assessments across pipeline actions" architecture rule)
 
 **Skip this step if the skill is running inside an AFK orchestrator** (e.g. `/wr-itil:work-problems`). Orchestrators handle release cadence themselves per the "Inter-iteration release cadence for AFK loops" architecture rule (Step 6.5). Detect via the presence of an orchestrator marker in the invoking prompt — look for phrases like "AFK", "work-problems", "batch-work", or the sentinel `ALL_DONE` convention. When in doubt, defer to the orchestrator by skipping this step.
 
 Otherwise, after the commit in step 11 lands, drain the release queue so the fix actually lands on npm without requiring manual user action.
 
-**Mechanism — delegate, do not re-implement scoring (per the "On-Demand Assessment Skills for Governance Plugins" architecture rule):**
+**Mechanism — reuse a valid cumulative assessment; delegate when needed (the "Reuse valid cumulative risk assessments across pipeline actions" architecture rule, the "On-Demand Assessment Skills for Governance Plugins" architecture rule):**
 
-1. Invoke the release scorer. Two paths are valid:
+1. Use the Step 11 pipeline assessment's separate `push` and `release` scores if it still covers this exact checkout, assessed content, unpushed and unreleased scope, and current effective appetite, and has not expired or drifted. A commit of identical assessed content does not require another scorer invocation. If any condition fails or cannot be established, invoke the scorer again before pushing or releasing. Two delegation paths are valid:
    - **Primary**: delegate to subagent type `wr-risk-scorer:pipeline` via the Agent tool.
    - **Fallback**: if that subagent type is not available, invoke skill `/wr-risk-scorer:assess-release` via the Skill tool. The skill wraps the same pipeline subagent.
-2. Read the returned `RISK_SCORES: commit=X push=Y release=Z` line.
-3. **Drain condition**: if `push` and `release` are both within appetite (≤ 4/25, "Low" band per `RISK-POLICY.md`), AND `.changeset/` is non-empty, proceed to the drain action. Otherwise, skip the drain and report the unreleased state.
+2. Read `RISK_SCORES: commit=X push=Y release=Z` from the valid assessment. Never substitute the commit score for a push or release score. The action-time `push:watch` and `release:watch` gates still check the relevant score, checkout, state, expiry, appetite, and CI; if a gate rejects stale or changed evidence, resolve the checkout binding or rescore before retrying.
+3. **Drain condition**: if `push` and `release` are both within appetite (at or below the effective appetite used by the scorer and gate), AND `.changeset/` is non-empty, proceed to the drain action. Otherwise, skip the drain and report the unreleased state.
 
 **Drain action (non-interactive, policy-authorised per the "Structured User Interaction for Governance-Skill Decisions" architecture rule Rule 6):**
 
@@ -1173,7 +1173,7 @@ Otherwise, after the commit in step 11 lands, drain the release queue so the fix
 
 **Failure handling**: If `release:watch` fails (CI failure, publish failure), stop and report the failure clearly. Do not retry non-interactively — the user must intervene.
 
-**Above-appetite branch (per the "Auto-apply scorer remediations to reach within appetite — open action-class vocabulary" architecture rule)**: If push or release risk is above appetite (≥ 5/25), the skill MUST auto-apply scorer remediations incrementally until residual risk converges within appetite, OR halt the skill per the "Auto-apply scorer remediations to reach within appetite — open action-class vocabulary" architecture rule Rule 5 if the scorer cannot produce a convergent plan. **The skill MUST NOT release above appetite under any circumstance.** The skill MUST NOT call `AskUserQuestion` as a shortcut out of the auto-apply loop.
+**Above-appetite branch (per the "Auto-apply scorer remediations to reach within appetite — open action-class vocabulary" architecture rule)**: If push or release risk is above appetite (above the effective appetite used by the scorer and gate), the skill MUST auto-apply scorer remediations incrementally until residual risk converges within appetite, OR halt the skill per the "Auto-apply scorer remediations to reach within appetite — open action-class vocabulary" architecture rule Rule 5 if the scorer cannot produce a convergent plan. **The skill MUST NOT release above appetite under any circumstance.** The skill MUST NOT call `AskUserQuestion` as a shortcut out of the auto-apply loop.
 
 **Auto-apply mechanism (the "Auto-apply scorer remediations to reach within appetite — open action-class vocabulary" architecture rule Rule 2):**
 
@@ -1196,7 +1196,7 @@ Otherwise, after the commit in step 11 lands, drain the release queue so the fix
 - The final `RISK_SCORES:` line
 - An "Auto-apply trail" subsection listing each remediation attempted with outcome
 - Any Verification Pending ticket IDs implicated per Rule 2b
-- A one-line scorer-gap note (e.g., "available remediations did not reduce residual risk below 5/25 — extend scorer vocabulary per the "Scorer `RISK_REMEDIATIONS:` block is free-form prose — agent must parse descriptions to decide what to do" problem")
+- A one-line scorer-gap note (e.g., "available remediations did not reduce residual risk to within the effective appetite — extend scorer vocabulary per the "Scorer `RISK_REMEDIATIONS:` block is free-form prose — agent must parse descriptions to decide what to do" problem")
 
 The user resolves interactively — typical resolutions include splitting the commit, feature-flagging the change, or opening a problem ticket documenting the scorer gap.
 

package/skills/mitigate-incident/SKILL.md

@@ -182,7 +182,7 @@ Otherwise, after the commit in step 8 lands, drain the release queue so the fix
    - **Primary**: delegate to subagent type `wr-risk-scorer:pipeline` via the Agent tool.
    - **Fallback**: if that subagent type is not available, invoke skill `/wr-risk-scorer:assess-release` via the Skill tool.
 2. Read the returned `RISK_SCORES: commit=X push=Y release=Z` line.
-3. **Drain condition**: if `push` and `release` are both within appetite (≤ 4/25, "Low" band per `RISK-POLICY.md`), AND `.changeset/` is non-empty, proceed to the drain action. Otherwise, skip the drain and report the unreleased state.
+3. **Drain condition**: if `push` and `release` are both within appetite (at or below the effective appetite used by the scorer and gate), AND `.changeset/` is non-empty, proceed to the drain action. Otherwise, skip the drain and report the unreleased state.
 
 **Drain action (non-interactive, policy-authorised per the "Structured User Interaction for Governance-Skill Decisions" architecture rule Rule 6):**
 

package/skills/restore-incident/SKILL.md

@@ -160,7 +160,7 @@ Otherwise, after the commit in step 8 lands, drain the release queue so the fix
    - **Primary**: delegate to subagent type `wr-risk-scorer:pipeline` via the Agent tool.
    - **Fallback**: if that subagent type is not available, invoke skill `/wr-risk-scorer:assess-release` via the Skill tool.
 2. Read the returned `RISK_SCORES: commit=X push=Y release=Z` line.
-3. **Drain condition**: if `push` and `release` are both within appetite (≤ 4/25, "Low" band per `RISK-POLICY.md`), AND `.changeset/` is non-empty, proceed to the drain action. Otherwise, skip the drain and report the unreleased state.
+3. **Drain condition**: if `push` and `release` are both within appetite (at or below the effective appetite used by the scorer and gate), AND `.changeset/` is non-empty, proceed to the drain action. Otherwise, skip the drain and report the unreleased state.
 
 **Drain action (non-interactive, policy-authorised per the "Structured User Interaction for Governance-Skill Decisions" architecture rule Rule 6):**
 

package/skills/update-upstream/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: wr-itil:update-upstream
-description: "Post a lifecycle-update comment to an upstream issue when a local problem ticket transitions. Drafts a transition-specific update (root-cause confirmed / fix released / closed), composes the prose through the external-comms risk gate + voice-tone gate, auto-posts within appetite, queues above-appetite. Reciprocal sibling to /wr-itil:report-upstream — initial-filing vs lifecycle-update split per the \"Cross-project problem-reporting contract — `report-upstream` skill in `@windyroad/itil`\" architecture rule amendment (the \"Problem 080: No bidirectional update of upstream-reported problems — local lifecycle transitions never propagate back to the reporter\" problem)."
+description: "Post a lifecycle-update comment to an upstream issue when a local problem ticket transitions. Drafts a transition-specific update (root-cause confirmed / fix released / closed), composes the prose through the external-comms risk gate + voice-tone gate, auto-posts on PASS, queues unresolved FAIL verdicts. Reciprocal sibling to /wr-itil:report-upstream — initial-filing vs lifecycle-update split per the \"Cross-project problem-reporting contract — `report-upstream` skill in `@windyroad/itil`\" architecture rule amendment (the \"Problem 080: No bidirectional update of upstream-reported problems — local lifecycle transitions never propagate back to the reporter\" problem)."
 allowed-tools: Read, Write, Edit, Bash, Glob, Grep, AskUserQuestion, Skill, Agent
 ---
 
@@ -23,7 +23,7 @@ allowed-tools: Read, Write, Edit, Bash, Glob, Grep, AskUserQuestion, Skill, Agen
 
 # Update Upstream — Lifecycle-Update Skill
 
-Post a lifecycle-update comment to an upstream issue or pull request when the local problem ticket transitions. Reads the local ticket's `## Reported Upstream` section, drafts a transition-specific update from the templates below, composes the draft through the external-comms risk gate (`wr-risk-scorer:external-comms`) and voice-tone gate (`wr-voice-tone:external-comms`), auto-posts via `gh issue comment` or `gh pr comment` when both gates pass within appetite, closes only issue targets on Verifying → Closed, and queues an `outstanding_questions` entry when either gate scores above appetite.
+Post a lifecycle-update comment to an upstream issue or pull request when the local problem ticket transitions. Reads the local ticket's `## Reported Upstream` section, drafts a transition-specific update from the templates below, composes the draft through the external-comms risk gate (`wr-risk-scorer:external-comms`) and voice-tone gate (`wr-voice-tone:external-comms`), auto-posts via `gh issue comment` or `gh pr comment` when both gates PASS, closes only issue targets on Verifying → Closed, and queues an `outstanding_questions` entry when either gate still FAILS after risk reduction.
 
 This skill is the **reciprocal sibling** to [`/wr-itil:report-upstream`](../report-upstream/SKILL.md) — that skill files the initial upstream report; this skill keeps the upstream record in sync as the local ticket walks its lifecycle. The split is per [the "Rename `wr-problem` Plugin to `wr-itil`" architecture rule](../../../docs/decisions/010-rename-wr-problem-to-wr-itil.proposed.md) amended Skill Granularity rule (one skill per distinct user intent) — initial-filing and lifecycle-update are distinct user intents with distinct autocomplete surfaces.
 
@@ -49,8 +49,8 @@ The single-ticket form is typically invoked from `/wr-itil:transition-problem` S
 - Determine the local ticket's current Status from the filename suffix.
 - Draft a transition-specific lifecycle-update comment per the templates below (Open→KE / KE→Verifying / Verifying→Closed).
 - Compose the drafted prose through `wr-risk-scorer:external-comms` + `wr-voice-tone:external-comms` gates.
-- Within appetite → post via `gh issue comment <n>`, or `gh pr comment <n>` when the disclosure path records a pull request (the "Prefer an upstream pull request over an issue when the upstream accepts pull requests" architecture rule); on Verifying→Closed also run `gh issue close <n>` — **but never `gh pr close`** (see below).
-- Above appetite → AskUserQuestion (interactive) / queue `outstanding_questions` (AFK, per the "AFK iter default when a skill needs to ask a question and AskUserQuestion is unavailable — should queue the question and move to the next iteration (not halt, not silently skip)" problem queue-and-continue).
+- Both gates PASS → post via `gh issue comment <n>`, or `gh pr comment <n>` when the disclosure path records a pull request (the "Prefer an upstream pull request over an issue when the upstream accepts pull requests" architecture rule); on Verifying→Closed also run `gh issue close <n>` — **but never `gh pr close`** (see below).
+- Gate FAIL → AskUserQuestion (interactive) / queue `outstanding_questions` (AFK, per the "AFK iter default when a skill needs to ask a question and AskUserQuestion is unavailable — should queue the question and move to the next iteration (not halt, not silently skip)" problem queue-and-continue).
 - Back-write a `## Upstream Lifecycle Updates` log entry to the local ticket recording the transition, the matched URL, the posted comment URL, and the disclosure path.
 - **Historical catch-up migration (`--catchup`, the "Problem 080: No bidirectional update of upstream-reported problems — local lifecycle transitions never propagate back to the reporter" problem Phase 2)** — one-shot retroactive scan of the existing `.verifying.md` + `.closed.md` corpus; posts the lifecycle update each linked-upstream ticket should already carry. Idempotent — re-running is safe. See [§ Catchup migration mode](#catchup-migration-mode-phase-2).
 
@@ -105,7 +105,7 @@ For each `- **URL**: <url>` line under `## Reported Upstream`, extract:
 
 Skip entries whose disclosure path is `drafted-and-saved (mailbox / out-of-band)` — those reports were never filed via `gh`, so there is no issue to comment on. Log a one-line skip note: `Skipping upstream entry <url> — disclosure path is out-of-band; user follow-up required.` The skip is **not** queued as an `outstanding_questions` entry (the user already owns the out-of-band channel per the "Cross-project problem-reporting contract — `report-upstream` skill in `@windyroad/itil`" architecture rule no-infra-for-email constraint).
 
-Multiple `## Reported Upstream` entries are supported (the local ticket may have been filed to multiple upstream trackers). Process each entry independently — the gate composition + post + back-write all run per-entry; one above-appetite entry queues only that entry, the rest proceed.
+Multiple `## Reported Upstream` entries are supported (the local ticket may have been filed to multiple upstream trackers). Process each entry independently — the gate composition + post + back-write all run per-entry; one failed gate queues only that entry, the rest proceed.
 
 ### 3. Determine the transition that fired this invocation
 
@@ -223,33 +223,36 @@ Score the drafted comment body via the `wr-risk-scorer:external-comms` agent (sh
 The agent returns a structured verdict:
 
 ```
-EXTERNAL_COMMS_RISK_VERDICT
-band: Low (<=4/25) | Medium (5..16) | High (17+)
-score: <0..25>
-pass: true | false
-reason: <one-line rationale>
+EXTERNAL_COMMS_RISK_VERDICT: PASS
 ```
 
-- **`pass: true` AND band ≤ Low (4/25)**: within appetite per RISK-POLICY.md commit-layer. Proceed to 5b (voice-tone).
-- **`pass: false` OR band > Low**: above appetite. Branch to 5c (above-appetite handling).
+Or, on failure:
+
+```
+EXTERNAL_COMMS_RISK_VERDICT: FAIL
+EXTERNAL_COMMS_RISK_REASON: <one-line rationale>
+```
+
+- **PASS**: proceed to 5b (voice-tone).
+- **FAIL**: branch to 5c (gate-failure handling).
 
 #### 5b. Voice-tone gate
 
 `gh issue comment` and `gh issue close` are on the "External-comms gate — voice-tone + risk/leak evaluators on shared PreToolUse surface" architecture rule gated surface list. The PreToolUse:Bash hook fires deny-plus-delegate to `wr-voice-tone:agent`. The agent reads the drafted body against `docs/VOICE-AND-TONE.md` and writes the bypass marker on PASS; the original `gh` call retries automatically.
 
-A FAIL verdict on the voice-tone gate is treated identically to an above-appetite risk verdict — branch to 5c.
+A FAIL verdict on the voice-tone gate also branches to 5c.
 
-#### 5c. Above-appetite handling
+#### 5c. Gate-failure handling
 
-The decision policy here is **framework-resolved** per [the "— Decision-Delegation Contract: when agents act on the framework vs ask the user" architecture rule](../../../docs/decisions/044-decision-delegation-contract.proposed.md) (decision-delegation contract) and the "Structured User Interaction for Governance-Skill Decisions" architecture rule Rule 6 (AFK fail-safe). No per-transition `AskUserQuestion` for the GATE FIRING — the gate scoring is itself the framework. The above-appetite handling differs by orchestrator context:
+The decision policy here is **framework-resolved** per [the "— Decision-Delegation Contract: when agents act on the framework vs ask the user" architecture rule](../../../docs/decisions/044-decision-delegation-contract.proposed.md) (decision-delegation contract) and the "Structured User Interaction for Governance-Skill Decisions" architecture rule Rule 6 (AFK fail-safe). No per-transition `AskUserQuestion` for the GATE FIRING — the gate verdict is itself the framework. The failure handling differs by orchestrator context:
 
 - **Interactive context** (per the "Structured User Interaction for Governance-Skill Decisions" architecture rule Rule 1): use `AskUserQuestion` to surface the drafted comment + the gate verdict + the matched URL, with options:
   - `Post the comment anyway (Recommended after review)` — user has read the draft and judged the post warranted; the skill bypasses the gate for this single post.
-  - `Risk-reduce and re-score` — invoke a tighter draft (shorter / fewer claims / stricter source-citation) and re-run the gate.
+  - `Risk-reduce and re-review` — invoke a tighter draft (shorter / fewer claims / stricter source-citation) and re-run the gate.
   - `Queue for later review` — save the draft to `## Queued Upstream Update` on the local ticket; user acts on return.
   - `Skip this update` — exit no-op for this upstream entry; the next transition's invocation re-considers.
 
-- **AFK / non-interactive context** (per the "Structured User Interaction for Governance-Skill Decisions" architecture rule Rule 6 + the "AFK iter default when a skill needs to ask a question and AskUserQuestion is unavailable — should queue the question and move to the next iteration (not halt, not silently skip)" problem queue-and-continue): the skill applies **silent risk-reduce + re-score** first — re-draft the comment with tighter source-citation + shorter prose, then re-invoke the external-comms gate. If the re-scored verdict is within appetite, proceed via 5b. Otherwise, save the drafted comment to the local ticket's `## Queued Upstream Update` section (shape below) and queue an `outstanding_questions` entry (category: `deviation-approval`) naming the local ticket ID + the matched URL + the residual band + the risk-reduce attempts taken. **The orchestrator continues per the "AFK iter default when a skill needs to ask a question and AskUserQuestion is unavailable — should queue the question and move to the next iteration (not halt, not silently skip)" problem** — do NOT halt the loop on an above-appetite upstream update.
+- **AFK / non-interactive context** (per the "Structured User Interaction for Governance-Skill Decisions" architecture rule Rule 6 + the "AFK iter default when a skill needs to ask a question and AskUserQuestion is unavailable — should queue the question and move to the next iteration (not halt, not silently skip)" problem queue-and-continue): the skill applies **silent risk-reduce + re-review** first — re-draft the comment with tighter source-citation + shorter prose, then re-invoke the external-comms gate. If the new verdict is PASS, proceed via 5b. Otherwise, save the drafted comment to the local ticket's `## Queued Upstream Update` section (shape below) and queue an `outstanding_questions` entry (category: `deviation-approval`) naming the local ticket ID + the matched URL + the FAIL reason + the risk-reduce attempts taken. **The orchestrator continues per the "AFK iter default when a skill needs to ask a question and AskUserQuestion is unavailable — should queue the question and move to the next iteration (not halt, not silently skip)" problem** — do NOT halt the loop on a failed upstream update gate.
 
 The silent risk-reduce step is **mechanical** per the "— Decision-Delegation Contract: when agents act on the framework vs ask the user" architecture rule framework-resolution boundary — the skill owns the re-draft; per-iter `AskUserQuestion` for risk-reduce vocabulary is the lazy-deferral anti-pattern the "Agents over-ask in interactive sessions — conflating mechanical-stages with user-interactive-stages of multi-stage skill contracts (inverse-)" problem closes.
 
@@ -261,8 +264,8 @@ The silent risk-reduce step is **mechanical** per the "— Decision-Delegation C
 - **Drafted**: <YYYY-MM-DD>
 - **Transition**: Open → Known Error | Known Error → Verification Pending | Verification Pending → Closed
 - **Target URL**: <upstream-issue-url>
-- **Halt reason**: above-appetite external-comms gate (band: <verdict band>; score: <verdict score>; reason: <verdict reason>) | above-appetite voice-tone gate (reason: <verdict reason>)
-- **Risk-reduce attempts**: <count, e.g. "1 — tighter source-citation; re-scored band Medium">
+- **Halt reason**: external-comms gate FAIL (reason: <EXTERNAL_COMMS_RISK_REASON>) | voice-tone gate FAIL (reason: <verdict reason>)
+- **Risk-reduce attempts**: <count, e.g. "1 — tighter source-citation; re-reviewed FAIL">
 - **Drafted comment body**:
 
   <the body that would have been posted as a `gh issue comment`, ready for manual review>
@@ -312,9 +315,9 @@ Append a log entry to the local ticket's `## Upstream Lifecycle Updates` section
 
 - **<YYYY-MM-DD>** — Open → Known Error
   - **Target URL**: <upstream-issue-url>
-  - **Comment URL**: <posted-comment-url> (or "queued — see ## Queued Upstream Update" when above-appetite)
-  - **Disclosure path**: posted-comment | posted-pr-comment (pull-request target — never closed) | posted-comment-and-closed (confirmed Verifying → Closed, issue targets only) | posted-comment-local-close-only (outbound evidence-authorised close) | queued-above-appetite | closed-already-upstream | skipped-out-of-band
-  - **Gate verdict**: external-comms <band/score> + voice-tone <pass|fail>
+  - **Comment URL**: <posted-comment-url> (or "queued — see ## Queued Upstream Update" when a gate fails)
+  - **Disclosure path**: posted-comment | posted-pr-comment (pull-request target — never closed) | posted-comment-and-closed (confirmed Verifying → Closed, issue targets only) | posted-comment-local-close-only (outbound evidence-authorised close) | queued-gate-failure | closed-already-upstream | skipped-out-of-band
+  - **Gate verdict**: external-comms <PASS|FAIL> + voice-tone <PASS|FAIL>
 
 - **<YYYY-MM-DD>** — Known Error → Verification Pending
   - ... (next entry appends; never replaces earlier entries)
@@ -494,7 +497,7 @@ Route the GENERATED comment through the reporter-facing gate chain — **cogniti
 
 **Cog-a11y gate (when-available, the "Phase 2 — cognitive-accessibility evaluator on the 4 external-comms surfaces (gh / npm / changeset / git commit), shipped as a NEW `@windyroad/cognitive-a11y` plugin" problem-gated — do NOT block):** the cognitive-accessibility evaluator rides FIRST so reporter-facing prose is checked for plain-language / reading-level before the risk + voice-tone legs. `@windyroad/cognitive-a11y` does not exist yet ([the "Phase 2 — cognitive-accessibility evaluator on the 4 external-comms surfaces (gh / npm / changeset / git commit), shipped as a NEW `@windyroad/cognitive-a11y` plugin" problem](../../../docs/problems/open/338-p082-phase-2-cognitive-a11y-evaluator-on-external-comms-surfaces-new-windyroad-cognitive-a11y-plugin.md) Open) — until it lands the chain degrades to the existing external-comms + voice-tone dual gate, exactly as the "External-comms gate — voice-tone + risk/leak evaluators on shared PreToolUse surface" architecture rule's per-evaluator marker scheme handles an uninstalled evaluator (an absent evaluator's gate is simply not registered → the remaining legs' PASS unblocks the retry). The cog-a11y-as-third-external-comms-evaluator declaration is recorded in [the "External-comms gate — voice-tone + risk/leak evaluators on shared PreToolUse surface" architecture rule](../../../docs/decisions/028-voice-tone-gate-external-comms.proposed.md)'s `## Amendments` (the locus the "External-comms gate — voice-tone + risk/leak evaluators on shared PreToolUse surface" architecture rule's own Reassessment Criteria designate for a third evaluator); this leg is the consumer-side wiring note only. **Do NOT block this iteration's inbound dispatch on the "Phase 2 — cognitive-accessibility evaluator on the 4 external-comms surfaces (gh / npm / changeset / git commit), shipped as a NEW `@windyroad/cognitive-a11y` plugin" problem** — ship the wiring, ride the dual gate today.
 
-Above-appetite handling is identical to Step 5c (silent risk-reduce + re-score; if still above, save to `## Queued Upstream Update` + queue an `outstanding_questions` entry; the orchestrator continues per the "AFK iter default when a skill needs to ask a question and AskUserQuestion is unavailable — should queue the question and move to the next iteration (not halt, not silently skip)" problem — do NOT halt). Within appetite, post on our own repo:
+Gate-failure handling is identical to Step 5c (silent risk-reduce + re-review; if still FAIL, save to `## Queued Upstream Update` + queue an `outstanding_questions` entry; the orchestrator continues per the "AFK iter default when a skill needs to ask a question and AskUserQuestion is unavailable — should queue the question and move to the next iteration (not halt, not silently skip)" problem — do NOT halt). When both gates PASS, post on our own repo:
 
 ```bash
 gh issue comment "${NN}" --repo "${OWN_OWNER_REPO}" --body "${INBOUND_BODY}"
@@ -517,16 +520,16 @@ Append to the same `## Upstream Lifecycle Updates` log (Step 6 shape), tagged fo
 ```markdown
 - **<YYYY-MM-DD>** — Known Error → Verification Pending (inbound)
   - **Target**: inbound #<NN> (own repo <OWN_OWNER_REPO>)
-  - **Comment URL**: <posted-comment-url> (or "skipped — already-posted-inbound" when the idempotency guard matched, or "queued — see ## Queued Upstream Update" when above-appetite)
-  - **Disclosure path**: posted-inbound-comment | posted-inbound-comment-and-closed (proven owned issue, confirmed or evidence-authorised close) | already-posted-inbound | queued-above-appetite | closed-already | inbound-channel-unresolved
-  - **Gate verdict**: external-comms <band/score> + voice-tone <pass|fail>
+  - **Comment URL**: <posted-comment-url> (or "skipped — already-posted-inbound" when the idempotency guard matched, or "queued — see ## Queued Upstream Update" when a gate fails)
+  - **Disclosure path**: posted-inbound-comment | posted-inbound-comment-and-closed (proven owned issue, confirmed or evidence-authorised close) | already-posted-inbound | queued-gate-failure | closed-already | inbound-channel-unresolved
+  - **Gate verdict**: external-comms <PASS|FAIL> + voice-tone <PASS|FAIL>
 ```
 
 The log stays append-only and direction-tagged; `/wr-itil:check-upstream-responses` (the "No process for issue reporters to check for responses — symmetric gap to inbound discovery" problem) does NOT read this section, so logging inbound entries here cannot contaminate the outbound poller.
 
 #### I7. Both-direction tickets
 
-A ticket may carry BOTH a `## Reported Upstream` section AND an inbound `**Origin**` field (we reported it upstream *and* someone reported it against us). The two legs are **independent**: the outbound entries run Steps 2–6 against their external URLs; the inbound leg runs I1–I6 against the own-repo issue. Each gate-composes, posts, and logs separately — one above-appetite leg queues only itself; the other proceeds.
+A ticket may carry BOTH a `## Reported Upstream` section AND an inbound `**Origin**` field (we reported it upstream *and* someone reported it against us). The two legs are **independent**: the outbound entries run Steps 2–6 against their external URLs; the inbound leg runs I1–I6 against the own-repo issue. Each gate-composes, posts, and logs separately — one failed gate queues only its own leg; the other proceeds.
 
 ### 7. Commit per the "Governance Skills Commit Their Own Completed Work" architecture rule
 
@@ -578,10 +581,10 @@ The append-only log (written by Step 6 on every post) is the source of truth —
 For each `CATCHUP` line, parse its `disclosure=` token and run the **existing per-ticket flow** (Steps 4–6) against that ticket ID. `disclosure=pull-request` selects `gh pr comment` and forbids any close command; `disclosure=issue` selects `gh issue comment` and permits `gh issue close` for Verifying → Closed:
 
 1. Draft the transition template (Step 4) for the entry's transition (`KE->Verifying` → Known Error → Verification Pending template; `Verifying->Closed` → Verification Pending → Closed template). Only an issue disclosure also runs `gh issue close`; a pull-request disclosure never runs `gh pr close`.
-2. Compose through the external-comms + voice-tone gates (Step 5) — **identical** dual-gate composition as the per-ticket path. Above-appetite handling (Step 5c) is unchanged: silent risk-reduce + re-score, then queue to `## Queued Upstream Update` + `outstanding_questions` (category `deviation-approval`) per the "AFK iter default when a skill needs to ask a question and AskUserQuestion is unavailable — should queue the question and move to the next iteration (not halt, not silently skip)" problem if still above. Catchup does NOT bypass the gates.
-3. Post within appetite (Step 5b final) and back-write the `## Upstream Lifecycle Updates` log (Step 6).
+2. Compose through the external-comms + voice-tone gates (Step 5) — **identical** dual-gate composition as the per-ticket path. Gate-failure handling (Step 5c) is unchanged: silent risk-reduce + re-review, then queue to `## Queued Upstream Update` + `outstanding_questions` (category `deviation-approval`) per the "AFK iter default when a skill needs to ask a question and AskUserQuestion is unavailable — should queue the question and move to the next iteration (not halt, not silently skip)" problem if still FAIL. Catchup does NOT bypass the gates.
+3. Post after both gates PASS (Step 5b final) and back-write the `## Upstream Lifecycle Updates` log (Step 6).
 
-Process entries one at a time so a single above-appetite entry queues only itself; the rest proceed. There is no batch-cap on the number of catchup posts — the gate composition is the rate-limit, and the corpus is bounded (one pass over local tickets).
+Process entries one at a time so a single failed gate queues only its own entry; the rest proceed. There is no batch-cap on the number of catchup posts — the gate composition is the rate-limit, and the corpus is bounded (one pass over local tickets).
 
 ### C4. Commit per the "Governance Skills Commit Their Own Completed Work" architecture rule
 
@@ -593,15 +596,15 @@ The live-upstream end-to-end confirmation (the "Problem 080: No bidirectional up
 
 ## AFK behaviour summary
 
-Four distinct AFK branches. Per the [the "Cross-project problem-reporting contract — `report-upstream` skill in `@windyroad/itil`" architecture rule](../../../docs/decisions/024-cross-project-problem-reporting-contract.proposed.md) amendment (the "Problem 080: No bidirectional update of upstream-reported problems — local lifecycle transitions never propagate back to the reporter" problem) — same composition shape as the post-the "Agent waits for human to initiate upstream report instead of filing on detect — feedback delay class" problem initial-filing path — ALL pre-post branches route through the `wr-risk-scorer:external-comms` + `wr-voice-tone:external-comms` gates. Below-appetite proceeds; above-appetite silent risk-reduces + re-scores; if still above, queues per the "AFK iter default when a skill needs to ask a question and AskUserQuestion is unavailable — should queue the question and move to the next iteration (not halt, not silently skip)" problem queue-and-continue without halting the loop.
+Four distinct AFK branches. Per the [the "Cross-project problem-reporting contract — `report-upstream` skill in `@windyroad/itil`" architecture rule](../../../docs/decisions/024-cross-project-problem-reporting-contract.proposed.md) amendment (the "Problem 080: No bidirectional update of upstream-reported problems — local lifecycle transitions never propagate back to the reporter" problem) — same composition shape as the post-the "Agent waits for human to initiate upstream report instead of filing on detect — feedback delay class" problem initial-filing path — ALL pre-post branches route through the `wr-risk-scorer:external-comms` + `wr-voice-tone:external-comms` gates. PASS proceeds; FAIL silently risk-reduces + re-reviews; an unresolved FAIL queues per the "AFK iter default when a skill needs to ask a question and AskUserQuestion is unavailable — should queue the question and move to the next iteration (not halt, not silently skip)" problem queue-and-continue without halting the loop.
 
 | Branch | AFK behaviour | Authority |
 |---|---|---|
-| Below-appetite post (Step 5b final) | Post via `gh issue comment`; on Verifying→Closed also `gh issue close`. Back-write to `## Upstream Lifecycle Updates`. Voice-tone gate per the "External-comms gate — voice-tone + risk/leak evaluators on shared PreToolUse surface" architecture rule may delegate-and-retry on the post; treat as expected. | the "Cross-project problem-reporting contract — `report-upstream` skill in `@windyroad/itil`" architecture rule amendment (the "Problem 080: No bidirectional update of upstream-reported problems — local lifecycle transitions never propagate back to the reporter" problem); the "External-comms gate — voice-tone + risk/leak evaluators on shared PreToolUse surface" architecture rule |
-| Above-appetite — silent risk-reduce + re-score within appetite | Re-draft with tighter source-citation + shorter prose; re-invoke `wr-risk-scorer:external-comms`. If within → post per the below-appetite branch. | the "Cross-project problem-reporting contract — `report-upstream` skill in `@windyroad/itil`" architecture rule amendment (the "Problem 080: No bidirectional update of upstream-reported problems — local lifecycle transitions never propagate back to the reporter" problem); the "— Decision-Delegation Contract: when agents act on the framework vs ask the user" architecture rule framework-resolution boundary; the "Auto-apply scorer remediations to reach within appetite — open action-class vocabulary" architecture rule within-axis precedent (open-vocabulary risk-reducing measures) |
-| Above-appetite — silent risk-reduce did not bring within appetite | Save drafted comment to `## Queued Upstream Update` + queue `outstanding_questions` entry (category: `deviation-approval`). Orchestrator continues per the "AFK iter default when a skill needs to ask a question and AskUserQuestion is unavailable — should queue the question and move to the next iteration (not halt, not silently skip)" problem. | the "Cross-project problem-reporting contract — `report-upstream` skill in `@windyroad/itil`" architecture rule amendment (the "Problem 080: No bidirectional update of upstream-reported problems — local lifecycle transitions never propagate back to the reporter" problem); the "Structured User Interaction for Governance-Skill Decisions" architecture rule Rule 6; the "AFK iter default when a skill needs to ask a question and AskUserQuestion is unavailable — should queue the question and move to the next iteration (not halt, not silently skip)" problem |
+| Both gates PASS (Step 5b final) | Post via `gh issue comment`; on Verifying→Closed also `gh issue close`. Back-write to `## Upstream Lifecycle Updates`. Voice-tone gate per the "External-comms gate — voice-tone + risk/leak evaluators on shared PreToolUse surface" architecture rule may delegate-and-retry on the post; treat as expected. | the "Cross-project problem-reporting contract — `report-upstream` skill in `@windyroad/itil`" architecture rule amendment (the "Problem 080: No bidirectional update of upstream-reported problems — local lifecycle transitions never propagate back to the reporter" problem); the "External-comms gate — voice-tone + risk/leak evaluators on shared PreToolUse surface" architecture rule |
+| Gate FAIL — silent risk-reduce yields PASS | Re-draft with tighter source-citation + shorter prose; re-invoke `wr-risk-scorer:external-comms`. On PASS → post through the voice-tone gate. | the "Cross-project problem-reporting contract — `report-upstream` skill in `@windyroad/itil`" architecture rule amendment (the "Problem 080: No bidirectional update of upstream-reported problems — local lifecycle transitions never propagate back to the reporter" problem); the "— Decision-Delegation Contract: when agents act on the framework vs ask the user" architecture rule framework-resolution boundary; the "Auto-apply scorer remediations to reach within appetite — open action-class vocabulary" architecture rule within-axis precedent (open-vocabulary risk-reducing measures) |
+| Gate FAIL — silent risk-reduce still FAILS | Save drafted comment to `## Queued Upstream Update` + queue `outstanding_questions` entry (category: `deviation-approval`). Orchestrator continues per the "AFK iter default when a skill needs to ask a question and AskUserQuestion is unavailable — should queue the question and move to the next iteration (not halt, not silently skip)" problem. | the "Cross-project problem-reporting contract — `report-upstream` skill in `@windyroad/itil`" architecture rule amendment (the "Problem 080: No bidirectional update of upstream-reported problems — local lifecycle transitions never propagate back to the reporter" problem); the "Structured User Interaction for Governance-Skill Decisions" architecture rule Rule 6; the "AFK iter default when a skill needs to ask a question and AskUserQuestion is unavailable — should queue the question and move to the next iteration (not halt, not silently skip)" problem |
 | Above-appetite commit (Step 7) | Skip the commit, report uncommitted state. | the "Structured User Interaction for Governance-Skill Decisions" architecture rule Rule 6 |
-| **Inbound-origin verdict (the "Inbound-reported tickets never receive fix-released verdict on originating issue" problem)** | When the ticket carries `**Origin**: inbound-reported (#NN)`, run the [§ Inbound-origin verdict dispatch](#inbound-origin-verdict-dispatch-p363) leg (I1–I7): idempotency-guard, then **GENERATE** the verdict comment from ticket context (no templates; O→KE shares the workaround with provenance-credit), route through the cog-a11y → external-comms → voice-tone gate chain (cog-a11y when-available per the "Phase 2 — cognitive-accessibility evaluator on the 4 external-comms surfaces (gh / npm / changeset / git commit), shipped as a NEW `@windyroad/cognitive-a11y` plugin" problem; dual gate today), post `gh issue comment` on our own repo (and `gh issue close` on Verifying → Closed), back-write a direction-tagged lifecycle log entry. Above-appetite queues per the rows above (does NOT halt). Reporter-facing prose; anti-leakage visibility-gated (PUBLIC → titled+linked refs; PRIVATE/indeterminate → strict ban; classification tokens / step IDs / internal vocab always banned — the "Inbound-discovery ack comments are bureaucratic, not verdict-shaped ( violation)" problem/the "agent surfaces decisions to user using opaque IDs (P-numbers, ADR-numbers, JTBD-numbers) without explaining what they mean — empathy gap" problem). | the "Cross-project problem-reporting contract — `report-upstream` skill in `@windyroad/itil`" architecture rule amendment (the "Inbound-reported tickets never receive fix-released verdict on originating issue" problem); the "External-comms gate — voice-tone + risk/leak evaluators on shared PreToolUse surface" architecture rule (cog-a11y third evaluator, the "Phase 2 — cognitive-accessibility evaluator on the 4 external-comms surfaces (gh / npm / changeset / git commit), shipped as a NEW `@windyroad/cognitive-a11y` plugin" problem); the "Plugin-published artefacts use namespace-prefixed permalinks for internal IDs" architecture rule (linked-title); the "Inbound-reported problems rank ahead of internally-discovered problems via a sort tier" architecture rule (Origin field); the "AFK iter default when a skill needs to ask a question and AskUserQuestion is unavailable — should queue the question and move to the next iteration (not halt, not silently skip)" problem |
+| **Inbound-origin verdict (the "Inbound-reported tickets never receive fix-released verdict on originating issue" problem)** | When the ticket carries `**Origin**: inbound-reported (#NN)`, run the [§ Inbound-origin verdict dispatch](#inbound-origin-verdict-dispatch-p363) leg (I1–I7): idempotency-guard, then **GENERATE** the verdict comment from ticket context (no templates; O→KE shares the workaround with provenance-credit), route through the cog-a11y → external-comms → voice-tone gate chain (cog-a11y when-available per the "Phase 2 — cognitive-accessibility evaluator on the 4 external-comms surfaces (gh / npm / changeset / git commit), shipped as a NEW `@windyroad/cognitive-a11y` plugin" problem; dual gate today), post `gh issue comment` on our own repo (and `gh issue close` on Verifying → Closed), back-write a direction-tagged lifecycle log entry. Failed gates queue per the rows above (does NOT halt). Reporter-facing prose; anti-leakage visibility-gated (PUBLIC → titled+linked refs; PRIVATE/indeterminate → strict ban; classification tokens / step IDs / internal vocab always banned — the "Inbound-discovery ack comments are bureaucratic, not verdict-shaped ( violation)" problem/the "agent surfaces decisions to user using opaque IDs (P-numbers, ADR-numbers, JTBD-numbers) without explaining what they mean — empathy gap" problem). | the "Cross-project problem-reporting contract — `report-upstream` skill in `@windyroad/itil`" architecture rule amendment (the "Inbound-reported tickets never receive fix-released verdict on originating issue" problem); the "External-comms gate — voice-tone + risk/leak evaluators on shared PreToolUse surface" architecture rule (cog-a11y third evaluator, the "Phase 2 — cognitive-accessibility evaluator on the 4 external-comms surfaces (gh / npm / changeset / git commit), shipped as a NEW `@windyroad/cognitive-a11y` plugin" problem); the "Plugin-published artefacts use namespace-prefixed permalinks for internal IDs" architecture rule (linked-title); the "Inbound-reported problems rank ahead of internally-discovered problems via a sort tier" architecture rule (Origin field); the "AFK iter default when a skill needs to ask a question and AskUserQuestion is unavailable — should queue the question and move to the next iteration (not halt, not silently skip)" problem |
 
 The pre-amendment "halt-the-orchestrator on above-appetite" semantics are **superseded** by queue-and-continue per the "AFK iter default when a skill needs to ask a question and AskUserQuestion is unavailable — should queue the question and move to the next iteration (not halt, not silently skip)" problem — same shape as the post-the "Agent waits for human to initiate upstream report instead of filing on detect — feedback delay class" problem initial-filing path.
 

package/skills/work-problems/SKILL.md

@@ -986,9 +986,10 @@ After the iteration's commit lands but before starting the next iteration, check
    - **Fallback**: if that subagent type is not available, invoke skill `/wr-risk-scorer:assess-release` via the Skill tool. The skill wraps the same pipeline subagent.
 2. Read the returned `RISK_SCORES: commit=X push=Y release=Z` line and the `RISK_REMEDIATIONS:` block (if present).
 3. **Classify the residual + queue state (the "work-problems Step 6.5 "≤3 within appetite — no drain" clause defers low-risk releases, encoding accumulation" problem)**:
-   - **Above appetite (≥ 5/25)** — route to the **Above-appetite branch** below. Do NOT drain. Do NOT proceed to Step 6.75 until either (a) the auto-apply loop re-converges within appetite and drain succeeds, or (b) Rule 5 halt fires.
-   - **Within appetite (≤ 4/25) AND there is releasable material** (any unpushed commits on `HEAD..origin/<base>` OR any entries in `.changeset/`) — drain the queue per the Drain action below, then proceed to Step 6.75. The release-action threshold is "is there something to release?", NOT "has accumulated risk reached the safety band?" Per user direction 2026-05-17 (the "work-problems Step 6.5 "≤3 within appetite — no drain" clause defers low-risk releases, encoding accumulation" problem Description): *"If it's low risk, you should release."* Low cost to release + low residual risk = release now; never accumulate.
-   - **Within appetite (≤ 4/25) AND empty queue** (no unpushed commits AND no `.changeset/` entries) — no drain (literally nothing to release). Proceed to Step 6.75. This is the genuine no-op fast-path; the gate is *absence of releasable material*, not residual band.
+   - Use the scorer's **effective appetite** (`RISK_APPETITE` override, then `RISK-POLICY.md`, then default 5); a score equal to that appetite is within it.
+   - **Above appetite (push or release score > effective appetite)** — route to the **Above-appetite branch** below. Do NOT drain. Do NOT proceed to Step 6.75 until either (a) the auto-apply loop re-converges within appetite and drain succeeds, or (b) Rule 5 halt fires.
+   - **Within appetite (both scores ≤ effective appetite) AND there is releasable material** (any unpushed commits on `HEAD..origin/<base>` OR any entries in `.changeset/`) — drain the queue per the Drain action below, then proceed to Step 6.75. The release-action threshold is "is there something to release?", NOT "has accumulated risk reached the safety band?" Per user direction 2026-05-17 (the "work-problems Step 6.5 "≤3 within appetite — no drain" clause defers low-risk releases, encoding accumulation" problem Description): *"If it's low risk, you should release."* Low cost to release + low residual risk = release now; never accumulate.
+   - **Within appetite (both scores ≤ effective appetite) AND empty queue** (no unpushed commits AND no `.changeset/` entries) — no drain (literally nothing to release). Proceed to Step 6.75. This is the genuine no-op fast-path; the gate is *absence of releasable material*, not residual band.
 
 **Drain action (non-interactive, policy-authorised per the "Structured User Interaction for Governance-Skill Decisions" architecture rule Rule 6):**
 
@@ -1068,7 +1069,7 @@ After a successful release-cadence drain has shipped a new plugin version to npm
 
 #### Above-appetite branch (per the "Auto-apply scorer remediations to reach within appetite — open action-class vocabulary" architecture rule)
 
-**Invariant**: the orchestrator MUST NOT release above appetite. There is no code path in Step 6.5 that releases at residual push/release ≥ 5/25. The orchestrator MUST NOT call `AskUserQuestion` as a shortcut out of the auto-apply loop — the scorer is the decision surface, not the user. The branch terminates in either a within-appetite drain or a Rule 5 halt.
+**Invariant**: the orchestrator MUST NOT release above the effective appetite. There is no code path in Step 6.5 that releases when either residual push or release score exceeds it. The orchestrator MUST NOT call `AskUserQuestion` as a shortcut out of the auto-apply loop — the scorer is the decision surface, not the user. The branch terminates in either a within-appetite drain or a Rule 5 halt.
 
 **Auto-apply loop (the "Auto-apply scorer remediations to reach within appetite — open action-class vocabulary" architecture rule Rule 2):**
 
@@ -1086,8 +1087,8 @@ After a successful release-cadence drain has shipped a new plugin version to npm
    - `disable-or-revert`: remove the risky behaviour from shipped code until its required evidence or control exists.
 5. Re-invoke the risk scorer (same delegation path as step 1 above — subagent preferred, skill fallback). Read the new `RISK_SCORES:` line.
 6. **Loop classification**:
-   - **Re-score within appetite (≤ 4/25)** — proceed to Drain action above. Done with the above-appetite branch.
-   - **Re-score still above appetite (≥ 5/25)** — continue working to reduce risk. The agent reads the new remediations and decides what to do next. Loop.
+   - **Re-score within appetite (both scores ≤ effective appetite)** — proceed to Drain action above. Done with the above-appetite branch.
+   - **Re-score still above appetite (either score > effective appetite)** — continue working to reduce risk. The agent reads the new remediations and decides what to do next. Loop.
    - **No remediations remain** or **the agent has exhausted its own ideas** — Rule 5 halt.
 
 **Governance gates per auto-apply (the "Auto-apply scorer remediations to reach within appetite — open action-class vocabulary" architecture rule Rule 3):** each auto-apply that requires a commit (the amend in step 4 above) goes through the standard the "Governance Skills Commit Their Own Completed Work" architecture rule commit flow — architect review, JTBD review, risk-scorer gate. A gate rejection falls through to Rule 5 halt. The scorer's suggestions do NOT bypass gates.
@@ -1159,12 +1160,12 @@ When `AskUserQuestion` is unavailable or the user is AFK, the skill (and the del
 | Scope expansion during work | Update problem file, re-score WSJF, move to next problem instead of continuing |
 | Commit when risk within appetite | Auto-commit (manage-problem step 9e fallback) |
 | Commit when risk above appetite | Skip commit, report uncommitted state |
-| Pipeline risk within appetite (≤ 4/25) with releasable material (any unpushed commits OR any `.changeset/` entries) | Drain release queue (`push:watch` then, if releasable changesets exist, `release:watch`) before next iteration — per the "Inter-iteration release cadence for AFK loops" architecture rule (Step 6.5) as amended by the "work-problems Step 6.5 "≤3 within appetite — no drain" clause defers low-risk releases, encoding accumulation" problem. Trigger is *presence of releasable material*, not residual band reaching appetite. User direction 2026-05-17: "If it's low risk, you should release." |
-| Pipeline risk within appetite (≤ 4/25) AND empty queue (no unpushed commits AND no `.changeset/`) | No drain — literally nothing to release. Proceed directly to Step 6.75. The genuine no-op fast-path per the "work-problems Step 6.5 "≤3 within appetite — no drain" clause defers low-risk releases, encoding accumulation" problem. |
+| Pipeline risk within effective appetite (both scores ≤ effective appetite) with releasable material (any unpushed commits OR any `.changeset/` entries) | Drain release queue (`push:watch` then, if releasable changesets exist, `release:watch`) before next iteration — per the "Inter-iteration release cadence for AFK loops" architecture rule (Step 6.5) as amended by the "work-problems Step 6.5 "≤3 within appetite — no drain" clause defers low-risk releases, encoding accumulation" problem. Trigger is *presence of releasable material*, not residual band reaching appetite. User direction 2026-05-17: "If it's low risk, you should release." |
+| Pipeline risk within effective appetite (both scores ≤ effective appetite) AND empty queue (no unpushed commits AND no `.changeset/`) | No drain — literally nothing to release. Proceed directly to Step 6.75. The genuine no-op fast-path per the "work-problems Step 6.5 "≤3 within appetite — no drain" clause defers low-risk releases, encoding accumulation" problem. |
 | Post-release K→V auto-transition between iters (the ".known-error.md → .verifying.md transition not happening consistently at release time" problem) | After a successful within-appetite Drain action shipped a release to npm, invoke `wr-itil-enumerate-postrelease-kv-candidates` to enumerate `.known-error.md` tickets whose `**Release vehicle**: .changeset/<name>.md` citation matches a just-shipped (deleted-from-tree) changeset, and dispatch `/wr-itil:transition-problem <NNN> verifying` per emitted `KV_CANDIDATE` line. Conditional on actual release (skipped when `push:watch` ran alone with no changeset); non-blocking on individual transition failure (logs per-ticket, continues to next candidate; persistent failures route to Step 2.5b accumulated questions). V→C remains a maintainer-only surface — this callback fires K→V only. Per the "Problem lifecycle — add a Verification Pending status between Known Error and Closed" architecture rule + the "Inter-iteration release cadence for AFK loops" architecture rule + the "Rename `wr-problem` Plugin to `wr-itil`" architecture rule amended the "Problem 093: `/wr-itil:transition-problem` ↔ `/wr-itil:manage-problem` circular delegation for `<NNN> <status>` args" problem + the "Governance Skills Commit Their Own Completed Work" architecture rule + the "Structured User Interaction for Governance-Skill Decisions" architecture rule Rule 5 + the "— Decision-Delegation Contract: when agents act on the framework vs ask the user" architecture rule + the ".known-error.md → .verifying.md transition not happening consistently at release time" problem + the "AFK iter subprocess plugin cache stale after release — just-shipped hook does not protect the next iter" problem + the "Codify `derive-release-vehicle.sh` helper for K→V release-cycle citation" problem + the "derive-release-vehicle helper requires pre-edit of ticket changeset reference — three-touch when one-touch would suffice" problem (Step 6.5 Post-release K→V auto-transition subsection). |
 | Post-release plugin cache refresh between iters (the "AFK iter subprocess plugin cache stale after release — just-shipped hook does not protect the next iter" problem) | After a successful within-appetite Drain action shipped a release to npm, chain `/install-updates` to refresh the plugin cache before the next iter dispatches. Conditional on actual release (skipped when `push:watch` ran alone with no changeset); non-blocking on `/install-updates` failure (degrades to cache-stays-stale, equivalent to pre-amendment behaviour). Mid-loop ask discipline preserved by treating any `/install-updates` AskUserQuestion surface AS the Non-interactive fallback dry-run path. Per the "Structured User Interaction for Governance-Skill Decisions" architecture rule Rule 5 + the "— Decision-Delegation Contract: when agents act on the framework vs ask the user" architecture rule + the "`/wr-itil:work-problems` orchestrator defaults to subprocess dispatch even when the user is observably interactive — loses real-time presence advantage" problem + the "`/install-updates` Step 7 uses `claude plugin install` which silently no-ops when a plugin is already installed — updates never actually land" problem + the "AFK iter subprocess plugin cache stale after release — just-shipped hook does not protect the next iter" problem (Step 6.5 Post-release cache refresh subsection). |
 | CI failure during Step 6.5 drain (within-appetite branch) | Diagnose via `gh run view --log-failed`, classify against the closed fixable-in-iter allow-list (the "Problem 081: Structural source-content tests are wasteful — TDD agent should reject them and require behavioural tests (+ framework / stub enhancements)" problem-class stale-grep-string, hook stub mismatch, test ID drift, environmental flake), fix-and-continue for fixable classes (each retry rides its own the "Governance Skills Commit Their Own Completed Work" architecture rule commit gate), 3-retry cap per iteration, halt for unrecoverable classes. Ambiguous classification defaults to halt. the "Structured User Interaction for Governance-Skill Decisions" architecture rule Rule 5 policy-authorised. Per the "Agent output grounding — no fabricated quantitative estimates, cite + persist + uncertainty" architecture rule grounding + the "— Decision-Delegation Contract: when agents act on the framework vs ask the user" architecture rule framework-resolution boundary + the "`/wr-itil:work-problems` Step 6.5 halt-on-CI-failure direction should be fix-and-continue when failure is mechanically fixable (-class stale assertions)" problem (Step 6.5 Failure handling). |
-| Pipeline risk above appetite (push or release >= 5/25) | Auto-apply scorer remediations incrementally (the "Auto-apply scorer remediations to reach within appetite — open action-class vocabulary" architecture rule Rule 2). The agent reads suggestions and decides what to do. Re-score after each apply; drain when within appetite. **Never release above appetite** (the "Auto-apply scorer remediations to reach within appetite — open action-class vocabulary" architecture rule Rule 1) — no AskUserQuestion shortcut. Halt the loop with `outcome: halted-above-appetite` if the loop exhausts without convergence (the "Auto-apply scorer remediations to reach within appetite — open action-class vocabulary" architecture rule Rule 5). Verification Pending commits excluded from auto-revert (Rule 2b). Per the "Auto-apply scorer remediations to reach within appetite — open action-class vocabulary" architecture rule (Step 6.5 Above-appetite branch). |
+| Pipeline risk above appetite (push or release score > effective appetite) | Auto-apply scorer remediations incrementally (the "Auto-apply scorer remediations to reach within appetite — open action-class vocabulary" architecture rule Rule 2). The agent reads suggestions and decides what to do. Re-score after each apply; drain when within appetite. **Never release above appetite** (the "Auto-apply scorer remediations to reach within appetite — open action-class vocabulary" architecture rule Rule 1) — no AskUserQuestion shortcut. Halt the loop with `outcome: halted-above-appetite` if the loop exhausts without convergence (the "Auto-apply scorer remediations to reach within appetite — open action-class vocabulary" architecture rule Rule 5). Verification Pending commits excluded from auto-revert (Rule 2b). Per the "Auto-apply scorer remediations to reach within appetite — open action-class vocabulary" architecture rule (Step 6.5 Above-appetite branch). |
 | Origin diverged before start (Branch 1) | Pull `--ff-only` if trivial; route to Branch 3 (stop with `git log HEAD..origin/<base>` and reverse report) if non-fast-forward — per the "AFK orchestrator preflight: get the repo into a clean state before starting" architecture rule (Step 0 Branch 1 / Branch 3). |
 | Pre-existing uncommitted work attributable to prior iter's in-flight flow (Branch 2 — DEFERRED) | Per the "AFK orchestrator preflight: get the repo into a clean state before starting" architecture rule Branch 2 (currently routes → Branch 3 until follow-up lands the auto-commit mechanism + the ": Enforce Governance Without Slowing Down" user outcome gate composition + bats). Auto-commit criteria when shipped: (a) provenance unambiguous AND (b) risk within appetite per the "Inter-iteration release cadence for AFK loops" architecture rule. Commit subject convention: `chore(preflight): recover prior-session in-flight work — <ticket-ref>` (the ": Progress the Backlog While I'm Away" user outcome audit trail). |
 | Prior-session partial work detected at start (Branch 3 detection — session-continuity dirty: untracked `docs/decisions/*.proposed.md` / `docs/problems/*.md`, `.afk-run-state/iter-*.json` with `is_error: true` or `api_error_status >= 400`, stale `.claude/worktrees/*`, uncommitted SKILL.md/source/ADR edits) | Halt the loop with a structured Prior-Session State report in the AFK summary — deliberate carve-out from the 2026-06-06 Rule 6 queue-and-continue default (ambiguous state would mask the bug this preflight surfaces). Do NOT attempt non-interactive resume. Interactive invocations prompt via `AskUserQuestion` with 4 options (resume / discard / leave-and-lower-priority / halt). Per the "`/wr-itil:work-problems` preflight Step 0 does not detect prior-session partial-work state (untracked ADRs, `.afk-run-state/iter-*.json` with 429/error statuses, existing `.claude/worktrees/*` branches)" problem + the "Structured User Interaction for Governance-Skill Decisions" architecture rule Rule 6 + the "AFK orchestrator preflight: get the repo into a clean state before starting" architecture rule (Step 0 Branch 3 detection mechanism). |

package/skills-codex/close-incident/SKILL.md

@@ -145,7 +145,7 @@ Otherwise, after the commit in step 7 lands, drain the release queue so the clos
    - **Primary**: delegate to subagent type `wr-risk-scorer:pipeline` via the native Codex subagent tool.
    - **Fallback**: if that subagent type is not available, invoke skill `/wr-risk-scorer:assess-release` via the installed skill invocation.
 2. Read the returned `RISK_SCORES: commit=X push=Y release=Z` line.
-3. **Drain condition**: if `push` and `release` are both within appetite (≤ 4/25, "Low" band per `RISK-POLICY.md`), AND `.changeset/` is non-empty, proceed to the drain action. Otherwise, skip the drain and report the unreleased state.
+3. **Drain condition**: if `push` and `release` are both within appetite (at or below the effective appetite used by the scorer and gate), AND `.changeset/` is non-empty, proceed to the drain action. Otherwise, skip the drain and report the unreleased state.
 
 **Drain action (non-interactive, policy-authorised per the "Structured User Interaction for Governance-Skill Decisions" architecture rule Rule 6):**
 

package/skills-codex/link-incident/SKILL.md

@@ -128,7 +128,7 @@ Otherwise, after the commit in step 7 lands, drain the release queue so the link
    - **Primary**: delegate to subagent type `wr-risk-scorer:pipeline` via the native Codex subagent tool.
    - **Fallback**: if that subagent type is not available, invoke skill `/wr-risk-scorer:assess-release` via the installed skill invocation.
 2. Read the returned `RISK_SCORES: commit=X push=Y release=Z` line.
-3. **Drain condition**: if `push` and `release` are both within appetite (≤ 4/25, "Low" band per `RISK-POLICY.md`), AND `.changeset/` is non-empty, proceed to the drain action. Otherwise, skip the drain and report the unreleased state.
+3. **Drain condition**: if `push` and `release` are both within appetite (at or below the effective appetite used by the scorer and gate), AND `.changeset/` is non-empty, proceed to the drain action. Otherwise, skip the drain and report the unreleased state.
 
 **Drain action (non-interactive, policy-authorised per the "Structured User Interaction for Governance-Skill Decisions" architecture rule Rule 6):**
 

package/skills-codex/manage-incident/SKILL.md

@@ -318,19 +318,19 @@ Commit the completed work per the "Governance Skills Commit Their Own Completed
    - Incident closed: `docs(incidents): close I<NNN>`
 4. If commit risk is above appetite: this is **framework-mediated, NOT a category-3 one-time-override ask** (the "Skills, agents, and hooks override RISK-POLICY appetite instead of applying it" problem/the ": Apply RISK-POLICY appetite faithfully across all surfaces" release design amendment 2026-06-24 — there is no incident carve-out). Per **the "Auto-apply scorer remediations to reach within appetite — open action-class vocabulary" architecture rule Rule 1b (incident-context scoring)**: an active incident is a risk being realised (Likelihood already 5), so the incident-response change is scored against that live realised-risk baseline — weighing P(the change increases impact) vs P(it reduces impact & likelihood / restores service) + P(it introduces a new incident). If net risk-reducing, it takes the risk-reducing path (the `reducing` bypass — `RISK_BYPASS: reducing`) and proceeds, no ask. This is *better* for restore-service-fast (the "Restore Service Fast with an Audit Trail" user outcome): a genuine hotfix clears via scoring with no consent gate mid-outage. If the change is NOT net-reducing, auto-remediate per the "Auto-apply scorer remediations to reach within appetite — open action-class vocabulary" architecture rule Rule 1 or halt per Rule 5. **MUST NOT commit above appetite; MUST NOT `request_user_input` "commit anyway".** If `request_user_input` is unavailable, the "Structured User Interaction for Governance-Skill Decisions" architecture rule Rule 6 fail-safe applies — skip the commit and report the uncommitted state.
 
-### 15. Auto-release when changesets are queued (the "Governance skills auto-release when changesets are queued" architecture rule)
+### 15. Auto-release when changesets are queued (the "Reuse valid cumulative risk assessments across pipeline actions" architecture rule)
 
 **Skip this step if the skill is running inside an AFK orchestrator.** Orchestrators handle release cadence themselves per the "Inter-iteration release cadence for AFK loops" architecture rule (Step 6.5). When in doubt, defer to the orchestrator by skipping this step.
 
 Otherwise, after the commit in step 14 lands, drain the release queue so the fix actually lands on npm without requiring manual user action.
 
-**Mechanism — delegate, do not re-implement scoring (per the "On-Demand Assessment Skills for Governance Plugins" architecture rule):**
+**Mechanism — reuse a valid cumulative assessment; delegate when needed (the "Reuse valid cumulative risk assessments across pipeline actions" architecture rule, the "On-Demand Assessment Skills for Governance Plugins" architecture rule):**
 
-1. Invoke the release scorer. Two paths are valid:
+1. Use the Step 14 pipeline assessment's separate `push` and `release` scores if it still covers this exact checkout, assessed content, unpushed and unreleased scope, and current effective appetite, and has not expired or drifted. A commit of identical assessed content does not require another scorer invocation. If any condition fails or cannot be established, invoke the scorer again before pushing or releasing. Two delegation paths are valid:
    - **Primary**: delegate to subagent type `wr-risk-scorer:pipeline` via the native Codex subagent tool.
    - **Fallback**: if that subagent type is not available, invoke skill `/wr-risk-scorer:assess-release` via the installed skill invocation. The skill wraps the same pipeline subagent.
-2. Read the returned `RISK_SCORES: commit=X push=Y release=Z` line.
-3. **Drain condition**: if `push` and `release` are both within appetite (≤ 4/25, "Low" band per `RISK-POLICY.md`), AND `.changeset/` is non-empty, proceed to the drain action. Otherwise, skip the drain and report the unreleased state.
+2. Read `RISK_SCORES: commit=X push=Y release=Z` from the valid assessment. Never substitute the commit score for a push or release score. The action-time `push:watch` and `release:watch` gates still check the relevant score, checkout, state, expiry, appetite, and CI; if a gate rejects stale or changed evidence, resolve the checkout binding or rescore before retrying.
+3. **Drain condition**: if `push` and `release` are both within appetite (at or below the effective appetite used by the scorer and gate), AND `.changeset/` is non-empty, proceed to the drain action. Otherwise, skip the drain and report the unreleased state.
 
 **Drain action (non-interactive, policy-authorised per the "Structured User Interaction for Governance-Skill Decisions" architecture rule Rule 6):**
 
@@ -340,7 +340,7 @@ Otherwise, after the commit in step 14 lands, drain the release queue so the fix
 
 **Failure handling**: If `release:watch` fails (CI failure, publish failure), stop and report the failure clearly. Do not retry non-interactively — the user must intervene.
 
-**Above-appetite branch (per the "Auto-apply scorer remediations to reach within appetite — open action-class vocabulary" architecture rule)**: If push or release risk is above appetite (≥ 5/25), the skill MUST auto-apply scorer remediations incrementally until residual risk converges within appetite, OR halt the skill per the "Auto-apply scorer remediations to reach within appetite — open action-class vocabulary" architecture rule Rule 5 if the scorer cannot produce a convergent plan. **The skill MUST NOT release above appetite under any circumstance.** The skill MUST NOT call `request_user_input` as a shortcut out of the auto-apply loop.
+**Above-appetite branch (per the "Auto-apply scorer remediations to reach within appetite — open action-class vocabulary" architecture rule)**: If push or release risk is above appetite (above the effective appetite used by the scorer and gate), the skill MUST auto-apply scorer remediations incrementally until residual risk converges within appetite, OR halt the skill per the "Auto-apply scorer remediations to reach within appetite — open action-class vocabulary" architecture rule Rule 5 if the scorer cannot produce a convergent plan. **The skill MUST NOT release above appetite under any circumstance.** The skill MUST NOT call `request_user_input` as a shortcut out of the auto-apply loop.
 
 **Auto-apply mechanism (the "Auto-apply scorer remediations to reach within appetite — open action-class vocabulary" architecture rule Rule 2):**
 
@@ -366,7 +366,7 @@ Otherwise, after the commit in step 14 lands, drain the release queue so the fix
 - **the "Add `manage-incident` Skill to `wr-itil` Plugin" architecture rule** (`docs/decisions/011-manage-incident-skill.proposed.md`) — incident lifecycle; evidence-first workflow; reversible-mitigation preference; Sev 4-5 lightweight path. Step 6's evidence-gate refactor (2026-04-28) extends the "Add `manage-incident` Skill to `wr-itil` Plugin" architecture rule's evidence-first rule with the documented `Record anyway` audit-trail bypass that mitigate-incident already used (cool-headed-commitment consistency across the two incident skills).
 - **the "Governance Skills Commit Their Own Completed Work" architecture rule** — governance skills commit their own work. Step 14 unchanged.
 - **the "On-Demand Assessment Skills for Governance Plugins" architecture rule** — release scorer delegation pattern. Step 15 unchanged.
-- **the "Inter-iteration release cadence for AFK loops" architecture rule** + **the "Governance skills auto-release when changesets are queued" architecture rule** — release cadence. Step 15 unchanged.
+- **the "Inter-iteration release cadence for AFK loops" architecture rule** + **the "Reuse valid cumulative risk assessments across pipeline actions" architecture rule** — AFK release cadence and attended reuse of valid cumulative risk scores in Step 15.
 - **the "Agent output grounding — no fabricated quantitative estimates, cite + persist + uncertainty" architecture rule** — cost-source grounding. Step 6's audit-trail bypass note preserves grounding by capturing the user's justification at deviation time.
 - **the "Auto-apply scorer remediations to reach within appetite — open action-class vocabulary" architecture rule** — auto-apply scorer remediations. Step 15 above-appetite branch unchanged.
 - **the "Problem 071: Argument-based skill subcommands are not discoverable in Codex autocomplete" problem** — skill-split origin (slice 6 — manage-incident is the host with thin-router forwarders for `list`, `mitigate`, `restored`, `close`, `link`).

package/skills-codex/manage-problem/SKILL.md

@@ -1158,23 +1158,23 @@ Commit the completed work per the "Governance Skills Commit Their Own Completed
    - Problem closed: `docs(problems): close P<NNN> <title>`
    - Review/re-rank: `docs(problems): review — re-rank priorities`
    - Fix implemented: `fix(<scope>): <description> (closes P<NNN>)` — include problem file changes (rename to `.verifying.md` + `## Fix Released` section) in the same commit per the "Problem lifecycle — add a Verification Pending status between Known Error and Closed" architecture rule
-5. If commit risk is above appetite: auto-apply scorer remediations per **the "Auto-apply scorer remediations to reach within appetite — open action-class vocabulary" architecture rule Rule 1** incrementally until residual commit risk is within appetite (≤ 4/25), OR halt per the "Auto-apply scorer remediations to reach within appetite — open action-class vocabulary" architecture rule Rule 5 if the scorer cannot converge. **The skill MUST NOT commit above appetite, and MUST NOT call `request_user_input` to ask whether to commit anyway** (the "Skills, agents, and hooks override RISK-POLICY appetite instead of applying it" problem/the ": Apply RISK-POLICY appetite faithfully across all surfaces" release design amendment 2026-06-24 — above-appetite is framework-mediated, never a category-3 one-time-override; same invariant the push/release branch at Step 12 already enforces). The "Structured User Interaction for Governance-Skill Decisions" architecture rule Rule 6 fail-safe (no `request_user_input` available / non-interactive → skip the commit and report the uncommitted state) remains the terminal fallback. This applies only to the risk-above-appetite branch, not to the delegation-unavailable case above.
+5. If commit risk is above appetite: auto-apply scorer remediations per **the "Auto-apply scorer remediations to reach within appetite — open action-class vocabulary" architecture rule Rule 1** incrementally until residual commit risk is within appetite (at or below the effective appetite used by the scorer and gate), OR halt per the "Auto-apply scorer remediations to reach within appetite — open action-class vocabulary" architecture rule Rule 5 if the scorer cannot converge. **The skill MUST NOT commit above appetite, and MUST NOT call `request_user_input` to ask whether to commit anyway** (the "Skills, agents, and hooks override RISK-POLICY appetite instead of applying it" problem/the ": Apply RISK-POLICY appetite faithfully across all surfaces" release design amendment 2026-06-24 — above-appetite is framework-mediated, never a category-3 one-time-override; same invariant the push/release branch at Step 12 already enforces). The "Structured User Interaction for Governance-Skill Decisions" architecture rule Rule 6 fail-safe (no `request_user_input` available / non-interactive → skip the commit and report the uncommitted state) remains the terminal fallback. This applies only to the risk-above-appetite branch, not to the delegation-unavailable case above.
 
 **Multi-commit slice changeset discipline (the "AFK iter `packages/<plugin>/` commits without changesets — orchestrator-main-turn back-fill is fragile recovery, hook-level enforcement preferable" problem Phase 2)**: when a single logical fix lands across multiple the "Governance Skills Commit Their Own Completed Work" architecture rule-grain commits targeting the same plugin (e.g. helper extraction in commit 1, callers wired in commit 2, SKILL note + transition in commit 3 — all `packages/<plugin>/`), author ONE changeset on the first commit in the slice. Subsequent same-plugin commits do NOT need their own changeset — the `itil-changeset-discipline.sh` hook's Check 2b recognises any `.changeset/*.md` already in the unpushed slice scope (`origin/<base>..HEAD` + untracked + modified-not-staged) that targets `"@windyroad/<plugin>": <any-bump>` and allows. This eliminates the per-commit changeset ceremony that previously produced N redundant `.changeset/*.md` files for one logical release entry (changesets-action collapses bump-class at version-package time, so per-commit changesets rendered N near-identical CHANGELOG bullets for one release). Once a changeset hits `origin/<base>` (drained at release time), it no longer counts — a fresh changeset is required for the next slice. Cross-plugin coverage is NOT permitted: an `@windyroad/itil` changeset does not satisfy a `packages/voice-tone/` commit.
 
-### 12. Auto-release when changesets are queued (the "Governance skills auto-release when changesets are queued" architecture rule)
+### 12. Auto-release when changesets are queued (the "Reuse valid cumulative risk assessments across pipeline actions" architecture rule)
 
 **Skip this step if the skill is running inside an AFK orchestrator** (e.g. `/wr-itil:work-problems`). Orchestrators handle release cadence themselves per the "Inter-iteration release cadence for AFK loops" architecture rule (Step 6.5). Detect via the presence of an orchestrator marker in the invoking prompt — look for phrases like "AFK", "work-problems", "batch-work", or the sentinel `ALL_DONE` convention. When in doubt, defer to the orchestrator by skipping this step.
 
 Otherwise, after the commit in step 11 lands, drain the release queue so the fix actually lands on npm without requiring manual user action.
 
-**Mechanism — delegate, do not re-implement scoring (per the "On-Demand Assessment Skills for Governance Plugins" architecture rule):**
+**Mechanism — reuse a valid cumulative assessment; delegate when needed (the "Reuse valid cumulative risk assessments across pipeline actions" architecture rule, the "On-Demand Assessment Skills for Governance Plugins" architecture rule):**
 
-1. Invoke the release scorer. Two paths are valid:
+1. Use the Step 11 pipeline assessment's separate `push` and `release` scores if it still covers this exact checkout, assessed content, unpushed and unreleased scope, and current effective appetite, and has not expired or drifted. A commit of identical assessed content does not require another scorer invocation. If any condition fails or cannot be established, invoke the scorer again before pushing or releasing. Two delegation paths are valid:
    - **Primary**: delegate to subagent type `wr-risk-scorer:pipeline` via the native Codex subagent tool.
    - **Fallback**: if that subagent type is not available, invoke skill `/wr-risk-scorer:assess-release` via the installed skill invocation. The skill wraps the same pipeline subagent.
-2. Read the returned `RISK_SCORES: commit=X push=Y release=Z` line.
-3. **Drain condition**: if `push` and `release` are both within appetite (≤ 4/25, "Low" band per `RISK-POLICY.md`), AND `.changeset/` is non-empty, proceed to the drain action. Otherwise, skip the drain and report the unreleased state.
+2. Read `RISK_SCORES: commit=X push=Y release=Z` from the valid assessment. Never substitute the commit score for a push or release score. The action-time `push:watch` and `release:watch` gates still check the relevant score, checkout, state, expiry, appetite, and CI; if a gate rejects stale or changed evidence, resolve the checkout binding or rescore before retrying.
+3. **Drain condition**: if `push` and `release` are both within appetite (at or below the effective appetite used by the scorer and gate), AND `.changeset/` is non-empty, proceed to the drain action. Otherwise, skip the drain and report the unreleased state.
 
 **Drain action (non-interactive, policy-authorised per the "Structured User Interaction for Governance-Skill Decisions" architecture rule Rule 6):**
 
@@ -1184,7 +1184,7 @@ Otherwise, after the commit in step 11 lands, drain the release queue so the fix
 
 **Failure handling**: If `release:watch` fails (CI failure, publish failure), stop and report the failure clearly. Do not retry non-interactively — the user must intervene.
 
-**Above-appetite branch (per the "Auto-apply scorer remediations to reach within appetite — open action-class vocabulary" architecture rule)**: If push or release risk is above appetite (≥ 5/25), the skill MUST auto-apply scorer remediations incrementally until residual risk converges within appetite, OR halt the skill per the "Auto-apply scorer remediations to reach within appetite — open action-class vocabulary" architecture rule Rule 5 if the scorer cannot produce a convergent plan. **The skill MUST NOT release above appetite under any circumstance.** The skill MUST NOT call `request_user_input` as a shortcut out of the auto-apply loop.
+**Above-appetite branch (per the "Auto-apply scorer remediations to reach within appetite — open action-class vocabulary" architecture rule)**: If push or release risk is above appetite (above the effective appetite used by the scorer and gate), the skill MUST auto-apply scorer remediations incrementally until residual risk converges within appetite, OR halt the skill per the "Auto-apply scorer remediations to reach within appetite — open action-class vocabulary" architecture rule Rule 5 if the scorer cannot produce a convergent plan. **The skill MUST NOT release above appetite under any circumstance.** The skill MUST NOT call `request_user_input` as a shortcut out of the auto-apply loop.
 
 **Auto-apply mechanism (the "Auto-apply scorer remediations to reach within appetite — open action-class vocabulary" architecture rule Rule 2):**
 
@@ -1207,7 +1207,7 @@ Otherwise, after the commit in step 11 lands, drain the release queue so the fix
 - The final `RISK_SCORES:` line
 - An "Auto-apply trail" subsection listing each remediation attempted with outcome
 - Any Verification Pending ticket IDs implicated per Rule 2b
-- A one-line scorer-gap note (e.g., "available remediations did not reduce residual risk below 5/25 — extend scorer vocabulary per the "Scorer `RISK_REMEDIATIONS:` block is free-form prose — agent must parse descriptions to decide what to do" problem")
+- A one-line scorer-gap note (e.g., "available remediations did not reduce residual risk to within the effective appetite — extend scorer vocabulary per the "Scorer `RISK_REMEDIATIONS:` block is free-form prose — agent must parse descriptions to decide what to do" problem")
 
 The user resolves interactively — typical resolutions include splitting the commit, feature-flagging the change, or opening a problem ticket documenting the scorer gap.
 

package/skills-codex/mitigate-incident/SKILL.md

@@ -193,7 +193,7 @@ Otherwise, after the commit in step 8 lands, drain the release queue so the fix
    - **Primary**: delegate to subagent type `wr-risk-scorer:pipeline` via the native Codex subagent tool.
    - **Fallback**: if that subagent type is not available, invoke skill `/wr-risk-scorer:assess-release` via the installed skill invocation.
 2. Read the returned `RISK_SCORES: commit=X push=Y release=Z` line.
-3. **Drain condition**: if `push` and `release` are both within appetite (≤ 4/25, "Low" band per `RISK-POLICY.md`), AND `.changeset/` is non-empty, proceed to the drain action. Otherwise, skip the drain and report the unreleased state.
+3. **Drain condition**: if `push` and `release` are both within appetite (at or below the effective appetite used by the scorer and gate), AND `.changeset/` is non-empty, proceed to the drain action. Otherwise, skip the drain and report the unreleased state.
 
 **Drain action (non-interactive, policy-authorised per the "Structured User Interaction for Governance-Skill Decisions" architecture rule Rule 6):**
 

package/skills-codex/restore-incident/SKILL.md

@@ -171,7 +171,7 @@ Otherwise, after the commit in step 8 lands, drain the release queue so the fix
    - **Primary**: delegate to subagent type `wr-risk-scorer:pipeline` via the native Codex subagent tool.
    - **Fallback**: if that subagent type is not available, invoke skill `/wr-risk-scorer:assess-release` via the installed skill invocation.
 2. Read the returned `RISK_SCORES: commit=X push=Y release=Z` line.
-3. **Drain condition**: if `push` and `release` are both within appetite (≤ 4/25, "Low" band per `RISK-POLICY.md`), AND `.changeset/` is non-empty, proceed to the drain action. Otherwise, skip the drain and report the unreleased state.
+3. **Drain condition**: if `push` and `release` are both within appetite (at or below the effective appetite used by the scorer and gate), AND `.changeset/` is non-empty, proceed to the drain action. Otherwise, skip the drain and report the unreleased state.
 
 **Drain action (non-interactive, policy-authorised per the "Structured User Interaction for Governance-Skill Decisions" architecture rule Rule 6):**
 

package/skills-codex/update-upstream/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: update-upstream
-description: "Post a lifecycle-update comment to an upstream issue when a local problem ticket transitions. Drafts a transition-specific update (root-cause confirmed / fix released / closed), composes the prose through the external-comms risk gate + voice-tone gate, auto-posts within appetite, queues above-appetite. Reciprocal sibling to /wr-itil:report-upstream — initial-filing vs lifecycle-update split per the \"Cross-project problem-reporting contract — `report-upstream` skill in `@windyroad/itil`\" architecture rule amendment (the \"Problem 080: No bidirectional update of upstream-reported problems — local lifecycle transitions never propagate back to the reporter\" problem)."
+description: "Post a lifecycle-update comment to an upstream issue when a local problem ticket transitions. Drafts a transition-specific update (root-cause confirmed / fix released / closed), composes the prose through the external-comms risk gate + voice-tone gate, auto-posts on PASS, queues unresolved FAIL verdicts. Reciprocal sibling to /wr-itil:report-upstream — initial-filing vs lifecycle-update split per the \"Cross-project problem-reporting contract — `report-upstream` skill in `@windyroad/itil`\" architecture rule amendment (the \"Problem 080: No bidirectional update of upstream-reported problems — local lifecycle transitions never propagate back to the reporter\" problem)."
 allowed-tools: Read, Write, Edit, Bash, Glob, Grep, request_user_input, Skill, Agent
 ---
 
@@ -34,7 +34,7 @@ allowed-tools: Read, Write, Edit, Bash, Glob, Grep, request_user_input, Skill, A
 
 # Update Upstream — Lifecycle-Update Skill
 
-Post a lifecycle-update comment to an upstream issue or pull request when the local problem ticket transitions. Reads the local ticket's `## Reported Upstream` section, drafts a transition-specific update from the templates below, composes the draft through the external-comms risk gate (`wr-risk-scorer:external-comms`) and voice-tone gate (`wr-voice-tone:external-comms`), auto-posts via `gh issue comment` or `gh pr comment` when both gates pass within appetite, closes only issue targets on Verifying → Closed, and queues an `outstanding_questions` entry when either gate scores above appetite.
+Post a lifecycle-update comment to an upstream issue or pull request when the local problem ticket transitions. Reads the local ticket's `## Reported Upstream` section, drafts a transition-specific update from the templates below, composes the draft through the external-comms risk gate (`wr-risk-scorer:external-comms`) and voice-tone gate (`wr-voice-tone:external-comms`), auto-posts via `gh issue comment` or `gh pr comment` when both gates PASS, closes only issue targets on Verifying → Closed, and queues an `outstanding_questions` entry when either gate still FAILS after risk reduction.
 
 This skill is the **reciprocal sibling** to [`/wr-itil:report-upstream`](../report-upstream/SKILL.md) — that skill files the initial upstream report; this skill keeps the upstream record in sync as the local ticket walks its lifecycle. The split is per [the "Rename `wr-problem` Plugin to `wr-itil`" architecture rule](../../../docs/decisions/010-rename-wr-problem-to-wr-itil.proposed.md) amended Skill Granularity rule (one skill per distinct user intent) — initial-filing and lifecycle-update are distinct user intents with distinct autocomplete surfaces.
 
@@ -60,8 +60,8 @@ The single-ticket form is typically invoked from `/wr-itil:transition-problem` S
 - Determine the local ticket's current Status from the filename suffix.
 - Draft a transition-specific lifecycle-update comment per the templates below (Open→KE / KE→Verifying / Verifying→Closed).
 - Compose the drafted prose through `wr-risk-scorer:external-comms` + `wr-voice-tone:external-comms` gates.
-- Within appetite → post via `gh issue comment <n>`, or `gh pr comment <n>` when the disclosure path records a pull request (the "Prefer an upstream pull request over an issue when the upstream accepts pull requests" architecture rule); on Verifying→Closed also run `gh issue close <n>` — **but never `gh pr close`** (see below).
-- Above appetite → request_user_input (interactive) / queue `outstanding_questions` (AFK, per the "AFK iter default when a skill needs to ask a question and request_user_input is unavailable — should queue the question and move to the next iteration (not halt, not silently skip)" problem queue-and-continue).
+- Both gates PASS → post via `gh issue comment <n>`, or `gh pr comment <n>` when the disclosure path records a pull request (the "Prefer an upstream pull request over an issue when the upstream accepts pull requests" architecture rule); on Verifying→Closed also run `gh issue close <n>` — **but never `gh pr close`** (see below).
+- Gate FAIL → request_user_input (interactive) / queue `outstanding_questions` (AFK, per the "AFK iter default when a skill needs to ask a question and request_user_input is unavailable — should queue the question and move to the next iteration (not halt, not silently skip)" problem queue-and-continue).
 - Back-write a `## Upstream Lifecycle Updates` log entry to the local ticket recording the transition, the matched URL, the posted comment URL, and the disclosure path.
 - **Historical catch-up migration (`--catchup`, the "Problem 080: No bidirectional update of upstream-reported problems — local lifecycle transitions never propagate back to the reporter" problem Phase 2)** — one-shot retroactive scan of the existing `.verifying.md` + `.closed.md` corpus; posts the lifecycle update each linked-upstream ticket should already carry. Idempotent — re-running is safe. See [§ Catchup migration mode](#catchup-migration-mode-phase-2).
 
@@ -116,7 +116,7 @@ For each `- **URL**: <url>` line under `## Reported Upstream`, extract:
 
 Skip entries whose disclosure path is `drafted-and-saved (mailbox / out-of-band)` — those reports were never filed via `gh`, so there is no issue to comment on. Log a one-line skip note: `Skipping upstream entry <url> — disclosure path is out-of-band; user follow-up required.` The skip is **not** queued as an `outstanding_questions` entry (the user already owns the out-of-band channel per the "Cross-project problem-reporting contract — `report-upstream` skill in `@windyroad/itil`" architecture rule no-infra-for-email constraint).
 
-Multiple `## Reported Upstream` entries are supported (the local ticket may have been filed to multiple upstream trackers). Process each entry independently — the gate composition + post + back-write all run per-entry; one above-appetite entry queues only that entry, the rest proceed.
+Multiple `## Reported Upstream` entries are supported (the local ticket may have been filed to multiple upstream trackers). Process each entry independently — the gate composition + post + back-write all run per-entry; one failed gate queues only that entry, the rest proceed.
 
 ### 3. Determine the transition that fired this invocation
 
@@ -234,33 +234,36 @@ Score the drafted comment body via the `wr-risk-scorer:external-comms` agent (sh
 The agent returns a structured verdict:
 
 ```
-EXTERNAL_COMMS_RISK_VERDICT
-band: Low (<=4/25) | Medium (5..16) | High (17+)
-score: <0..25>
-pass: true | false
-reason: <one-line rationale>
+EXTERNAL_COMMS_RISK_VERDICT: PASS
 ```
 
-- **`pass: true` AND band ≤ Low (4/25)**: within appetite per RISK-POLICY.md commit-layer. Proceed to 5b (voice-tone).
-- **`pass: false` OR band > Low**: above appetite. Branch to 5c (above-appetite handling).
+Or, on failure:
+
+```
+EXTERNAL_COMMS_RISK_VERDICT: FAIL
+EXTERNAL_COMMS_RISK_REASON: <one-line rationale>
+```
+
+- **PASS**: proceed to 5b (voice-tone).
+- **FAIL**: branch to 5c (gate-failure handling).
 
 #### 5b. Voice-tone gate
 
 `gh issue comment` and `gh issue close` are on the "External-comms gate — voice-tone + risk/leak evaluators on shared PreToolUse surface" architecture rule gated surface list. The PreToolUse:Bash hook fires deny-plus-delegate to `wr-voice-tone:agent`. The agent reads the drafted body against `docs/VOICE-AND-TONE.md` and writes the bypass marker on PASS; the original `gh` call retries automatically.
 
-A FAIL verdict on the voice-tone gate is treated identically to an above-appetite risk verdict — branch to 5c.
+A FAIL verdict on the voice-tone gate also branches to 5c.
 
-#### 5c. Above-appetite handling
+#### 5c. Gate-failure handling
 
-The decision policy here is **framework-resolved** per [the "— Decision-Delegation Contract: when agents act on the framework vs ask the user" architecture rule](../../../docs/decisions/044-decision-delegation-contract.proposed.md) (decision-delegation contract) and the "Structured User Interaction for Governance-Skill Decisions" architecture rule Rule 6 (AFK fail-safe). No per-transition `request_user_input` for the GATE FIRING — the gate scoring is itself the framework. The above-appetite handling differs by orchestrator context:
+The decision policy here is **framework-resolved** per [the "— Decision-Delegation Contract: when agents act on the framework vs ask the user" architecture rule](../../../docs/decisions/044-decision-delegation-contract.proposed.md) (decision-delegation contract) and the "Structured User Interaction for Governance-Skill Decisions" architecture rule Rule 6 (AFK fail-safe). No per-transition `request_user_input` for the GATE FIRING — the gate verdict is itself the framework. The failure handling differs by orchestrator context:
 
 - **Interactive context** (per the "Structured User Interaction for Governance-Skill Decisions" architecture rule Rule 1): use `request_user_input` to surface the drafted comment + the gate verdict + the matched URL, with options:
   - `Post the comment anyway (Recommended after review)` — user has read the draft and judged the post warranted; the skill bypasses the gate for this single post.
-  - `Risk-reduce and re-score` — invoke a tighter draft (shorter / fewer claims / stricter source-citation) and re-run the gate.
+  - `Risk-reduce and re-review` — invoke a tighter draft (shorter / fewer claims / stricter source-citation) and re-run the gate.
   - `Queue for later review` — save the draft to `## Queued Upstream Update` on the local ticket; user acts on return.
   - `Skip this update` — exit no-op for this upstream entry; the next transition's invocation re-considers.
 
-- **AFK / non-interactive context** (per the "Structured User Interaction for Governance-Skill Decisions" architecture rule Rule 6 + the "AFK iter default when a skill needs to ask a question and request_user_input is unavailable — should queue the question and move to the next iteration (not halt, not silently skip)" problem queue-and-continue): the skill applies **silent risk-reduce + re-score** first — re-draft the comment with tighter source-citation + shorter prose, then re-invoke the external-comms gate. If the re-scored verdict is within appetite, proceed via 5b. Otherwise, save the drafted comment to the local ticket's `## Queued Upstream Update` section (shape below) and queue an `outstanding_questions` entry (category: `deviation-approval`) naming the local ticket ID + the matched URL + the residual band + the risk-reduce attempts taken. **The orchestrator continues per the "AFK iter default when a skill needs to ask a question and request_user_input is unavailable — should queue the question and move to the next iteration (not halt, not silently skip)" problem** — do NOT halt the loop on an above-appetite upstream update.
+- **AFK / non-interactive context** (per the "Structured User Interaction for Governance-Skill Decisions" architecture rule Rule 6 + the "AFK iter default when a skill needs to ask a question and request_user_input is unavailable — should queue the question and move to the next iteration (not halt, not silently skip)" problem queue-and-continue): the skill applies **silent risk-reduce + re-review** first — re-draft the comment with tighter source-citation + shorter prose, then re-invoke the external-comms gate. If the new verdict is PASS, proceed via 5b. Otherwise, save the drafted comment to the local ticket's `## Queued Upstream Update` section (shape below) and queue an `outstanding_questions` entry (category: `deviation-approval`) naming the local ticket ID + the matched URL + the FAIL reason + the risk-reduce attempts taken. **The orchestrator continues per the "AFK iter default when a skill needs to ask a question and request_user_input is unavailable — should queue the question and move to the next iteration (not halt, not silently skip)" problem** — do NOT halt the loop on a failed upstream update gate.
 
 The silent risk-reduce step is **mechanical** per the "— Decision-Delegation Contract: when agents act on the framework vs ask the user" architecture rule framework-resolution boundary — the skill owns the re-draft; per-iter `request_user_input` for risk-reduce vocabulary is the lazy-deferral anti-pattern the "Agents over-ask in interactive sessions — conflating mechanical-stages with user-interactive-stages of multi-stage skill contracts (inverse-)" problem closes.
 
@@ -272,8 +275,8 @@ The silent risk-reduce step is **mechanical** per the "— Decision-Delegation C
 - **Drafted**: <YYYY-MM-DD>
 - **Transition**: Open → Known Error | Known Error → Verification Pending | Verification Pending → Closed
 - **Target URL**: <upstream-issue-url>
-- **Halt reason**: above-appetite external-comms gate (band: <verdict band>; score: <verdict score>; reason: <verdict reason>) | above-appetite voice-tone gate (reason: <verdict reason>)
-- **Risk-reduce attempts**: <count, e.g. "1 — tighter source-citation; re-scored band Medium">
+- **Halt reason**: external-comms gate FAIL (reason: <EXTERNAL_COMMS_RISK_REASON>) | voice-tone gate FAIL (reason: <verdict reason>)
+- **Risk-reduce attempts**: <count, e.g. "1 — tighter source-citation; re-reviewed FAIL">
 - **Drafted comment body**:
 
   <the body that would have been posted as a `gh issue comment`, ready for manual review>
@@ -323,9 +326,9 @@ Append a log entry to the local ticket's `## Upstream Lifecycle Updates` section
 
 - **<YYYY-MM-DD>** — Open → Known Error
   - **Target URL**: <upstream-issue-url>
-  - **Comment URL**: <posted-comment-url> (or "queued — see ## Queued Upstream Update" when above-appetite)
-  - **Disclosure path**: posted-comment | posted-pr-comment (pull-request target — never closed) | posted-comment-and-closed (confirmed Verifying → Closed, issue targets only) | posted-comment-local-close-only (outbound evidence-authorised close) | queued-above-appetite | closed-already-upstream | skipped-out-of-band
-  - **Gate verdict**: external-comms <band/score> + voice-tone <pass|fail>
+  - **Comment URL**: <posted-comment-url> (or "queued — see ## Queued Upstream Update" when a gate fails)
+  - **Disclosure path**: posted-comment | posted-pr-comment (pull-request target — never closed) | posted-comment-and-closed (confirmed Verifying → Closed, issue targets only) | posted-comment-local-close-only (outbound evidence-authorised close) | queued-gate-failure | closed-already-upstream | skipped-out-of-band
+  - **Gate verdict**: external-comms <PASS|FAIL> + voice-tone <PASS|FAIL>
 
 - **<YYYY-MM-DD>** — Known Error → Verification Pending
   - ... (next entry appends; never replaces earlier entries)
@@ -505,7 +508,7 @@ Route the GENERATED comment through the reporter-facing gate chain — **cogniti
 
 **Cog-a11y gate (when-available, the "Phase 2 — cognitive-accessibility evaluator on the 4 external-comms surfaces (gh / npm / changeset / git commit), shipped as a NEW `@windyroad/cognitive-a11y` plugin" problem-gated — do NOT block):** the cognitive-accessibility evaluator rides FIRST so reporter-facing prose is checked for plain-language / reading-level before the risk + voice-tone legs. `@windyroad/cognitive-a11y` does not exist yet ([the "Phase 2 — cognitive-accessibility evaluator on the 4 external-comms surfaces (gh / npm / changeset / git commit), shipped as a NEW `@windyroad/cognitive-a11y` plugin" problem](../../../docs/problems/open/338-p082-phase-2-cognitive-a11y-evaluator-on-external-comms-surfaces-new-windyroad-cognitive-a11y-plugin.md) Open) — until it lands the chain degrades to the existing external-comms + voice-tone dual gate, exactly as the "External-comms gate — voice-tone + risk/leak evaluators on shared PreToolUse surface" architecture rule's per-evaluator marker scheme handles an uninstalled evaluator (an absent evaluator's gate is simply not registered → the remaining legs' PASS unblocks the retry). The cog-a11y-as-third-external-comms-evaluator declaration is recorded in [the "External-comms gate — voice-tone + risk/leak evaluators on shared PreToolUse surface" architecture rule](../../../docs/decisions/028-voice-tone-gate-external-comms.proposed.md)'s `## Amendments` (the locus the "External-comms gate — voice-tone + risk/leak evaluators on shared PreToolUse surface" architecture rule's own Reassessment Criteria designate for a third evaluator); this leg is the consumer-side wiring note only. **Do NOT block this iteration's inbound dispatch on the "Phase 2 — cognitive-accessibility evaluator on the 4 external-comms surfaces (gh / npm / changeset / git commit), shipped as a NEW `@windyroad/cognitive-a11y` plugin" problem** — ship the wiring, ride the dual gate today.
 
-Above-appetite handling is identical to Step 5c (silent risk-reduce + re-score; if still above, save to `## Queued Upstream Update` + queue an `outstanding_questions` entry; the orchestrator continues per the "AFK iter default when a skill needs to ask a question and request_user_input is unavailable — should queue the question and move to the next iteration (not halt, not silently skip)" problem — do NOT halt). Within appetite, post on our own repo:
+Gate-failure handling is identical to Step 5c (silent risk-reduce + re-review; if still FAIL, save to `## Queued Upstream Update` + queue an `outstanding_questions` entry; the orchestrator continues per the "AFK iter default when a skill needs to ask a question and request_user_input is unavailable — should queue the question and move to the next iteration (not halt, not silently skip)" problem — do NOT halt). When both gates PASS, post on our own repo:
 
 ```bash
 gh issue comment "${NN}" --repo "${OWN_OWNER_REPO}" --body "${INBOUND_BODY}"
@@ -528,16 +531,16 @@ Append to the same `## Upstream Lifecycle Updates` log (Step 6 shape), tagged fo
 ```markdown
 - **<YYYY-MM-DD>** — Known Error → Verification Pending (inbound)
   - **Target**: inbound #<NN> (own repo <OWN_OWNER_REPO>)
-  - **Comment URL**: <posted-comment-url> (or "skipped — already-posted-inbound" when the idempotency guard matched, or "queued — see ## Queued Upstream Update" when above-appetite)
-  - **Disclosure path**: posted-inbound-comment | posted-inbound-comment-and-closed (proven owned issue, confirmed or evidence-authorised close) | already-posted-inbound | queued-above-appetite | closed-already | inbound-channel-unresolved
-  - **Gate verdict**: external-comms <band/score> + voice-tone <pass|fail>
+  - **Comment URL**: <posted-comment-url> (or "skipped — already-posted-inbound" when the idempotency guard matched, or "queued — see ## Queued Upstream Update" when a gate fails)
+  - **Disclosure path**: posted-inbound-comment | posted-inbound-comment-and-closed (proven owned issue, confirmed or evidence-authorised close) | already-posted-inbound | queued-gate-failure | closed-already | inbound-channel-unresolved
+  - **Gate verdict**: external-comms <PASS|FAIL> + voice-tone <PASS|FAIL>
 ```
 
 The log stays append-only and direction-tagged; `/wr-itil:check-upstream-responses` (the "No process for issue reporters to check for responses — symmetric gap to inbound discovery" problem) does NOT read this section, so logging inbound entries here cannot contaminate the outbound poller.
 
 #### I7. Both-direction tickets
 
-A ticket may carry BOTH a `## Reported Upstream` section AND an inbound `**Origin**` field (we reported it upstream *and* someone reported it against us). The two legs are **independent**: the outbound entries run Steps 2–6 against their external URLs; the inbound leg runs I1–I6 against the own-repo issue. Each gate-composes, posts, and logs separately — one above-appetite leg queues only itself; the other proceeds.
+A ticket may carry BOTH a `## Reported Upstream` section AND an inbound `**Origin**` field (we reported it upstream *and* someone reported it against us). The two legs are **independent**: the outbound entries run Steps 2–6 against their external URLs; the inbound leg runs I1–I6 against the own-repo issue. Each gate-composes, posts, and logs separately — one failed gate queues only its own leg; the other proceeds.
 
 ### 7. Commit per the "Governance Skills Commit Their Own Completed Work" architecture rule
 
@@ -589,10 +592,10 @@ The append-only log (written by Step 6 on every post) is the source of truth —
 For each `CATCHUP` line, parse its `disclosure=` token and run the **existing per-ticket flow** (Steps 4–6) against that ticket ID. `disclosure=pull-request` selects `gh pr comment` and forbids any close command; `disclosure=issue` selects `gh issue comment` and permits `gh issue close` for Verifying → Closed:
 
 1. Draft the transition template (Step 4) for the entry's transition (`KE->Verifying` → Known Error → Verification Pending template; `Verifying->Closed` → Verification Pending → Closed template). Only an issue disclosure also runs `gh issue close`; a pull-request disclosure never runs `gh pr close`.
-2. Compose through the external-comms + voice-tone gates (Step 5) — **identical** dual-gate composition as the per-ticket path. Above-appetite handling (Step 5c) is unchanged: silent risk-reduce + re-score, then queue to `## Queued Upstream Update` + `outstanding_questions` (category `deviation-approval`) per the "AFK iter default when a skill needs to ask a question and request_user_input is unavailable — should queue the question and move to the next iteration (not halt, not silently skip)" problem if still above. Catchup does NOT bypass the gates.
-3. Post within appetite (Step 5b final) and back-write the `## Upstream Lifecycle Updates` log (Step 6).
+2. Compose through the external-comms + voice-tone gates (Step 5) — **identical** dual-gate composition as the per-ticket path. Gate-failure handling (Step 5c) is unchanged: silent risk-reduce + re-review, then queue to `## Queued Upstream Update` + `outstanding_questions` (category `deviation-approval`) per the "AFK iter default when a skill needs to ask a question and request_user_input is unavailable — should queue the question and move to the next iteration (not halt, not silently skip)" problem if still FAIL. Catchup does NOT bypass the gates.
+3. Post after both gates PASS (Step 5b final) and back-write the `## Upstream Lifecycle Updates` log (Step 6).
 
-Process entries one at a time so a single above-appetite entry queues only itself; the rest proceed. There is no batch-cap on the number of catchup posts — the gate composition is the rate-limit, and the corpus is bounded (one pass over local tickets).
+Process entries one at a time so a single failed gate queues only its own entry; the rest proceed. There is no batch-cap on the number of catchup posts — the gate composition is the rate-limit, and the corpus is bounded (one pass over local tickets).
 
 ### C4. Commit per the "Governance Skills Commit Their Own Completed Work" architecture rule
 
@@ -604,15 +607,15 @@ The live-upstream end-to-end confirmation (the "Problem 080: No bidirectional up
 
 ## AFK behaviour summary
 
-Four distinct AFK branches. Per the [the "Cross-project problem-reporting contract — `report-upstream` skill in `@windyroad/itil`" architecture rule](../../../docs/decisions/024-cross-project-problem-reporting-contract.proposed.md) amendment (the "Problem 080: No bidirectional update of upstream-reported problems — local lifecycle transitions never propagate back to the reporter" problem) — same composition shape as the post-the "Agent waits for human to initiate upstream report instead of filing on detect — feedback delay class" problem initial-filing path — ALL pre-post branches route through the `wr-risk-scorer:external-comms` + `wr-voice-tone:external-comms` gates. Below-appetite proceeds; above-appetite silent risk-reduces + re-scores; if still above, queues per the "AFK iter default when a skill needs to ask a question and request_user_input is unavailable — should queue the question and move to the next iteration (not halt, not silently skip)" problem queue-and-continue without halting the loop.
+Four distinct AFK branches. Per the [the "Cross-project problem-reporting contract — `report-upstream` skill in `@windyroad/itil`" architecture rule](../../../docs/decisions/024-cross-project-problem-reporting-contract.proposed.md) amendment (the "Problem 080: No bidirectional update of upstream-reported problems — local lifecycle transitions never propagate back to the reporter" problem) — same composition shape as the post-the "Agent waits for human to initiate upstream report instead of filing on detect — feedback delay class" problem initial-filing path — ALL pre-post branches route through the `wr-risk-scorer:external-comms` + `wr-voice-tone:external-comms` gates. PASS proceeds; FAIL silently risk-reduces + re-reviews; an unresolved FAIL queues per the "AFK iter default when a skill needs to ask a question and request_user_input is unavailable — should queue the question and move to the next iteration (not halt, not silently skip)" problem queue-and-continue without halting the loop.
 
 | Branch | AFK behaviour | Authority |
 |---|---|---|
-| Below-appetite post (Step 5b final) | Post via `gh issue comment`; on Verifying→Closed also `gh issue close`. Back-write to `## Upstream Lifecycle Updates`. Voice-tone gate per the "External-comms gate — voice-tone + risk/leak evaluators on shared PreToolUse surface" architecture rule may delegate-and-retry on the post; treat as expected. | the "Cross-project problem-reporting contract — `report-upstream` skill in `@windyroad/itil`" architecture rule amendment (the "Problem 080: No bidirectional update of upstream-reported problems — local lifecycle transitions never propagate back to the reporter" problem); the "External-comms gate — voice-tone + risk/leak evaluators on shared PreToolUse surface" architecture rule |
-| Above-appetite — silent risk-reduce + re-score within appetite | Re-draft with tighter source-citation + shorter prose; re-invoke `wr-risk-scorer:external-comms`. If within → post per the below-appetite branch. | the "Cross-project problem-reporting contract — `report-upstream` skill in `@windyroad/itil`" architecture rule amendment (the "Problem 080: No bidirectional update of upstream-reported problems — local lifecycle transitions never propagate back to the reporter" problem); the "— Decision-Delegation Contract: when agents act on the framework vs ask the user" architecture rule framework-resolution boundary; the "Auto-apply scorer remediations to reach within appetite — open action-class vocabulary" architecture rule within-axis precedent (open-vocabulary risk-reducing measures) |
-| Above-appetite — silent risk-reduce did not bring within appetite | Save drafted comment to `## Queued Upstream Update` + queue `outstanding_questions` entry (category: `deviation-approval`). Orchestrator continues per the "AFK iter default when a skill needs to ask a question and request_user_input is unavailable — should queue the question and move to the next iteration (not halt, not silently skip)" problem. | the "Cross-project problem-reporting contract — `report-upstream` skill in `@windyroad/itil`" architecture rule amendment (the "Problem 080: No bidirectional update of upstream-reported problems — local lifecycle transitions never propagate back to the reporter" problem); the "Structured User Interaction for Governance-Skill Decisions" architecture rule Rule 6; the "AFK iter default when a skill needs to ask a question and request_user_input is unavailable — should queue the question and move to the next iteration (not halt, not silently skip)" problem |
+| Both gates PASS (Step 5b final) | Post via `gh issue comment`; on Verifying→Closed also `gh issue close`. Back-write to `## Upstream Lifecycle Updates`. Voice-tone gate per the "External-comms gate — voice-tone + risk/leak evaluators on shared PreToolUse surface" architecture rule may delegate-and-retry on the post; treat as expected. | the "Cross-project problem-reporting contract — `report-upstream` skill in `@windyroad/itil`" architecture rule amendment (the "Problem 080: No bidirectional update of upstream-reported problems — local lifecycle transitions never propagate back to the reporter" problem); the "External-comms gate — voice-tone + risk/leak evaluators on shared PreToolUse surface" architecture rule |
+| Gate FAIL — silent risk-reduce yields PASS | Re-draft with tighter source-citation + shorter prose; re-invoke `wr-risk-scorer:external-comms`. On PASS → post through the voice-tone gate. | the "Cross-project problem-reporting contract — `report-upstream` skill in `@windyroad/itil`" architecture rule amendment (the "Problem 080: No bidirectional update of upstream-reported problems — local lifecycle transitions never propagate back to the reporter" problem); the "— Decision-Delegation Contract: when agents act on the framework vs ask the user" architecture rule framework-resolution boundary; the "Auto-apply scorer remediations to reach within appetite — open action-class vocabulary" architecture rule within-axis precedent (open-vocabulary risk-reducing measures) |
+| Gate FAIL — silent risk-reduce still FAILS | Save drafted comment to `## Queued Upstream Update` + queue `outstanding_questions` entry (category: `deviation-approval`). Orchestrator continues per the "AFK iter default when a skill needs to ask a question and request_user_input is unavailable — should queue the question and move to the next iteration (not halt, not silently skip)" problem. | the "Cross-project problem-reporting contract — `report-upstream` skill in `@windyroad/itil`" architecture rule amendment (the "Problem 080: No bidirectional update of upstream-reported problems — local lifecycle transitions never propagate back to the reporter" problem); the "Structured User Interaction for Governance-Skill Decisions" architecture rule Rule 6; the "AFK iter default when a skill needs to ask a question and request_user_input is unavailable — should queue the question and move to the next iteration (not halt, not silently skip)" problem |
 | Above-appetite commit (Step 7) | Skip the commit, report uncommitted state. | the "Structured User Interaction for Governance-Skill Decisions" architecture rule Rule 6 |
-| **Inbound-origin verdict (the "Inbound-reported tickets never receive fix-released verdict on originating issue" problem)** | When the ticket carries `**Origin**: inbound-reported (#NN)`, run the [§ Inbound-origin verdict dispatch](#inbound-origin-verdict-dispatch-p363) leg (I1–I7): idempotency-guard, then **GENERATE** the verdict comment from ticket context (no templates; O→KE shares the workaround with provenance-credit), route through the cog-a11y → external-comms → voice-tone gate chain (cog-a11y when-available per the "Phase 2 — cognitive-accessibility evaluator on the 4 external-comms surfaces (gh / npm / changeset / git commit), shipped as a NEW `@windyroad/cognitive-a11y` plugin" problem; dual gate today), post `gh issue comment` on our own repo (and `gh issue close` on Verifying → Closed), back-write a direction-tagged lifecycle log entry. Above-appetite queues per the rows above (does NOT halt). Reporter-facing prose; anti-leakage visibility-gated (PUBLIC → titled+linked refs; PRIVATE/indeterminate → strict ban; classification tokens / step IDs / internal vocab always banned — the "Inbound-discovery ack comments are bureaucratic, not verdict-shaped ( violation)" problem/the "agent surfaces decisions to user using opaque IDs (P-numbers, ADR-numbers, JTBD-numbers) without explaining what they mean — empathy gap" problem). | the "Cross-project problem-reporting contract — `report-upstream` skill in `@windyroad/itil`" architecture rule amendment (the "Inbound-reported tickets never receive fix-released verdict on originating issue" problem); the "External-comms gate — voice-tone + risk/leak evaluators on shared PreToolUse surface" architecture rule (cog-a11y third evaluator, the "Phase 2 — cognitive-accessibility evaluator on the 4 external-comms surfaces (gh / npm / changeset / git commit), shipped as a NEW `@windyroad/cognitive-a11y` plugin" problem); the "Plugin-published artefacts use namespace-prefixed permalinks for internal IDs" architecture rule (linked-title); the "Inbound-reported problems rank ahead of internally-discovered problems via a sort tier" architecture rule (Origin field); the "AFK iter default when a skill needs to ask a question and request_user_input is unavailable — should queue the question and move to the next iteration (not halt, not silently skip)" problem |
+| **Inbound-origin verdict (the "Inbound-reported tickets never receive fix-released verdict on originating issue" problem)** | When the ticket carries `**Origin**: inbound-reported (#NN)`, run the [§ Inbound-origin verdict dispatch](#inbound-origin-verdict-dispatch-p363) leg (I1–I7): idempotency-guard, then **GENERATE** the verdict comment from ticket context (no templates; O→KE shares the workaround with provenance-credit), route through the cog-a11y → external-comms → voice-tone gate chain (cog-a11y when-available per the "Phase 2 — cognitive-accessibility evaluator on the 4 external-comms surfaces (gh / npm / changeset / git commit), shipped as a NEW `@windyroad/cognitive-a11y` plugin" problem; dual gate today), post `gh issue comment` on our own repo (and `gh issue close` on Verifying → Closed), back-write a direction-tagged lifecycle log entry. Failed gates queue per the rows above (does NOT halt). Reporter-facing prose; anti-leakage visibility-gated (PUBLIC → titled+linked refs; PRIVATE/indeterminate → strict ban; classification tokens / step IDs / internal vocab always banned — the "Inbound-discovery ack comments are bureaucratic, not verdict-shaped ( violation)" problem/the "agent surfaces decisions to user using opaque IDs (P-numbers, ADR-numbers, JTBD-numbers) without explaining what they mean — empathy gap" problem). | the "Cross-project problem-reporting contract — `report-upstream` skill in `@windyroad/itil`" architecture rule amendment (the "Inbound-reported tickets never receive fix-released verdict on originating issue" problem); the "External-comms gate — voice-tone + risk/leak evaluators on shared PreToolUse surface" architecture rule (cog-a11y third evaluator, the "Phase 2 — cognitive-accessibility evaluator on the 4 external-comms surfaces (gh / npm / changeset / git commit), shipped as a NEW `@windyroad/cognitive-a11y` plugin" problem); the "Plugin-published artefacts use namespace-prefixed permalinks for internal IDs" architecture rule (linked-title); the "Inbound-reported problems rank ahead of internally-discovered problems via a sort tier" architecture rule (Origin field); the "AFK iter default when a skill needs to ask a question and request_user_input is unavailable — should queue the question and move to the next iteration (not halt, not silently skip)" problem |
 
 The pre-amendment "halt-the-orchestrator on above-appetite" semantics are **superseded** by queue-and-continue per the "AFK iter default when a skill needs to ask a question and request_user_input is unavailable — should queue the question and move to the next iteration (not halt, not silently skip)" problem — same shape as the post-the "Agent waits for human to initiate upstream report instead of filing on detect — feedback delay class" problem initial-filing path.