@garygentry/feature-forge 0.3.5 → 0.3.7

package/adapters/claude/skills/forge-verify/SKILL.md

@@ -13,21 +13,21 @@ Analyze feature artifacts for completeness, consistency, and quality. Produce st
 
 This skill is loaded in two different roles. Determine yours before proceeding:
 
-- **You ARE the `forge-verifier` subagent** — you were dispatched via the Agent tool, you have read-only tools (Read, Glob, Grep, Bash) and **no** Agent/Task tool, and this skill is pre-loaded in your context. **SKIP "Subagent Delegation (parent orchestrator only)" and "Synthesize" below — those describe how a *parent* dispatches *you*, not work for you to do.** Do **not** dispatch anything, do **not** try to spawn a verifier. Go straight to **Prerequisites → Steps 1–6**, execute the checks yourself, and **return your findings as your response** (the parent writes the document to disk). Dispatching a subagent from here is the classic self-referential loop — never do it.
-- **You are the parent orchestrator** — a navigator (`/feature-forge:forge`), a stage skill's in-stage auto-verify, or a direct `/feature-forge:forge-verify` invocation, and you have the Agent tool. Use "Subagent Delegation" to dispatch the `forge-verifier` subagent, then "Synthesize" to assemble and write the document.
+- **You ARE the `forge-verifier` subagent** — you were dispatched via the `Agent` tool, you have read-only tools (Read, Glob, Grep, Bash) and **no** Agent/`Agent` tool, and this skill is pre-loaded in your context. **SKIP "Subagent Delegation (parent orchestrator only)" and "Synthesize" below — those describe how a *parent* dispatches *you*, not work for you to do.** Do **not** dispatch anything, do **not** try to spawn a verifier. Go straight to **Prerequisites → Steps 1–6**, execute the checks yourself, and **return your findings as your response** (the parent writes the document to disk). Dispatching a subagent from here is the classic self-referential loop — never do it.
+- **You are the parent orchestrator** — a navigator (`/feature-forge:forge`), a stage skill's in-stage auto-verify, or a direct `/feature-forge:forge-verify` invocation, and you have the `Agent` tool. Use "Subagent Delegation" to dispatch the `forge-verifier` subagent, then "Synthesize" to assemble and write the document.
 
 ## Subagent Delegation (parent orchestrator only)
 
-This skill is delegated to the `forge-verifier` subagent via the Agent tool. The verifier subagent has:
+This skill is delegated to the `forge-verifier` subagent via the `Agent` tool. The verifier subagent has:
 - **Read-only tools** (Read, Glob, Grep, Bash) — it cannot accidentally modify specs
-- **Persistent memory** — it accumulates knowledge about this project's recurring issues and patterns across sessions
+- **Persistent memory + shipped patterns** — it accumulates knowledge about this project's recurring issues across sessions, and on start reads the versioned, reviewed heuristics in `references/verifier-patterns/MEMORY.md` that ship with the pipeline (the two layers are distinct)
 - **The forge-verify skill pre-loaded** — so it has all verification checklists and guidance at startup
 
 ### Choose single vs. parallel dispatch
 
 Pick based on how many checks the mode carries (see the per-mode totals in Step 3):
 
-- **Small modes (prd 15, tech 17): single verifier.** Use the Agent tool once with
+- **Small modes (prd 15, tech 17): single verifier.** Use the `Agent` tool once with
   `subagent_type="forge-verifier"`, passing the feature name and mode. It runs all
   checks and returns findings.
 - **Large modes (specs 39, backlog 29, impl 25): parallel dimensioned fan-out.**
@@ -112,7 +112,7 @@ Resolve the feature directory via the **Feature Directory Resolution** block in
 
 Determine branch ownership **at entry**, from the literal `owner: nested` / `owner: direct` token in the dispatching prompt (absent the token you are `direct`), and preserve that value unchanged through any re-verify — see Step 7, which passes it through as `--owner`.
 
-**Turn structure reminder:** Output analysis/context as text, then route ALL questions through `AskUserQuestion`. Never embed questions in text output — the user will not be prompted and the session will stall.
+**Turn structure reminder:** Text first, then `AskUserQuestion` for all questions — never embed one in your text (it stalls). At rung 2/3, see the Interaction Capability Ladder.
 
 ## Step 1: Read Configuration and Determine Mode
 
@@ -170,7 +170,7 @@ Read `references/verification-checklists/{mode}.md` for the detailed checklist f
 
 Each check in that mode checklist has a unique ID (CHECK-P01, CHECK-T01, CHECK-S01, CHECK-B01, etc.). As you execute each check, record its ID and result (pass/fail/not-applicable). After completing all checks, report the total: "Executed N of M checks. Results: X pass, Y fail, Z not-applicable." If your count is significantly below the expected total for the mode (prd: 15 checks, tech: 17 checks, specs: 39 checks, backlog: 29 checks, impl: 25 checks, epic: 10 checks), you likely skipped checks — go back and complete them.
 
-**Epic mode dispatch.** Epic mode is a small (10-check) checklist, so per the single-vs-parallel rule above, dispatch a **single `forge-verifier`** via the Agent tool, passing the epic name and `mode=epic`. The verifier runs CHECK-E01..E10 from the `## Epic Mode Checklist` in `references/verification-checklists/epic.md` (E01/E02/E03/E08 are delegated to `epic-manifest.py validate`/`check-name`; E04–E07, E09, and E10 are verifier judgment) and returns its findings.
+**Epic mode dispatch.** Epic mode is a small (10-check) checklist, so per the single-vs-parallel rule above, dispatch a **single `forge-verifier`** via the `Agent` tool, passing the epic name and `mode=epic`. The verifier runs CHECK-E01..E10 from the `## Epic Mode Checklist` in `references/verification-checklists/epic.md` (E01/E02/E03/E08 are delegated to `epic-manifest.py validate`/`check-name`; E04–E07, E09, and E10 are verifier judgment) and returns its findings.
 
 ### Important: Be Specific, Not General
 
@@ -228,21 +228,23 @@ Do NOT embed this question in your text output.
 
 ## Step 6: Record the Result Through `state-verify`
 
-Never hand-author a verify entry. Every `stages.forge-verify-*` transition is written by the `state-verify` verb described in the **Pipeline State Protocol** in `references/shared-conventions.md`, which owns its full flag surface, its status matrix, and the exit-2 failure protocol. Write `findings-reported` when the report lists at least one **blocking** finding (`error`/`gap`); write `passed` when it lists none — including an **advisory-only** report (`inconsistency`/`improvement` findings only), which records `passed` with the report still attached: pass `--findings-file` and `--findings-count` alongside `--status passed` so the advisories remain discoverable without blocking the stage. (One exception routes blocking findings to `passed`: residual findings the user explicitly accepted at the round-ledger escalation — recorded first as a `state-decision`, then `passed` with the report attached, per "Escalation" in `references/stage-exit-protocol.md`.) Never write `findings-applied` here — that belongs to the fix pass. `--stage` names the **served production stage** (Step 1's mode, mapped through the served-stage mapping in Step 7), `--findings-file` is the report path **relative to** the feature directory (the Step 4 filename, round discriminator included), and `--verified-stage-version` is that production stage entry's current `version`, so a later revision of the artifact makes this verification read stale and re-fires. Add `--epic "{epic}"` when the feature is an epic member — required, per the Pipeline State Protocol; omitting it for a member is an error and must never fall back to a same-named flat feature.
+Never hand-author a verify entry. Every `stages.forge-verify-*` transition is written by the `state-verify` verb described in the **Pipeline State Protocol** in `references/shared-conventions.md`, which owns its full flag surface, its status matrix, and the exit-2 failure protocol. Write `findings-reported` when the report lists at least one **blocking** finding (`error`/`gap`); otherwise write `passed`. **Always attach the Step 4 report**, including a clean report with count `0`, so the completed round remains resumable and auditable. Advisory-only reports (`inconsistency`/`improvement`) use `passed` with their positive count and advance without a fix round. (One exception routes blocking findings to `passed`: residual findings the user explicitly accepted at the round-ledger escalation — recorded first as a `state-decision`, then `passed` with the report attached, per "Escalation" in `references/stage-exit-protocol.md`.) Never write `findings-applied` here — that belongs to the fix pass. `--stage` names the **served production stage** (Step 1's mode, mapped through the served-stage mapping in Step 7), `--findings-file` is the report path **relative to** the feature directory (the Step 4 filename, round discriminator included), and `--verified-stage-version` is that production stage entry's current `version`, so a later revision of the artifact makes this verification read stale and re-fires. Add `--epic "{epic}"` when the feature is an epic member — required, per the Pipeline State Protocol; omitting it for a member is an error and must never fall back to a same-named flat feature.
+
+Choose result arguments from the report outcome, never a combined placeholder: clean uses `--status passed --findings-count 0`; advisory-only uses `--status passed` with its positive count; blocking uses `--status findings-reported` with its positive total count. All variants attach the Step 4 report:
 
 ```bash
-R="$(bash -c 'for d in "${CLAUDE_PLUGIN_ROOT:-}" "$HOME"/.claude/skills/feature-forge "$HOME"/.claude/plugins/cache/*/feature-forge/* "$HOME"/.claude/plugins/*/feature-forge "$HOME"/.agents/skills/feature-forge ./.agents/skills/feature-forge; do [ -x "$d/scripts/forge-root.sh" ] && exec "$d/scripts/forge-root.sh"; done')"
+R="$(bash -c '[ -z "${FEATURE_FORGE_ROOT:-}" ] || [ -x "$FEATURE_FORGE_ROOT/scripts/forge-root.sh" ] || { echo "feature-forge: FEATURE_FORGE_ROOT=$FEATURE_FORGE_ROOT has no scripts/forge-root.sh" >&2; exit 2; }; for d in "${FEATURE_FORGE_ROOT:-}" "${CLAUDE_PLUGIN_ROOT:-}" "$HOME"/.claude/skills/feature-forge "$HOME"/.claude/plugins/cache/*/feature-forge/* "$HOME"/.claude/plugins/*/feature-forge "$HOME"/.agents/skills/feature-forge ./.agents/skills/feature-forge; do [ -x "$d/scripts/forge-root.sh" ] && exec "$d/scripts/forge-root.sh"; done')"
 [ -n "$R" ] || { echo "feature-forge: cannot locate plugin root" >&2; exit 1; }
 python3 "$R/scripts/forge-session.py" state-verify \
-  --feature "{feature}" --stage "{servedStage}" --status "{passed|findings-reported}" \
-  --findings-file "{relative findings path}" --findings-count {n} \
+  --feature "{feature}" --stage "{servedStage}" --status "{outcome-specific status}" \
+  --findings-file "{relative findings path}" --findings-count {outcome-specific count} \
   --verified-stage-version {version} --specs-dir "{specsDir}"
 ```
 
 **Two-commit provenance — never `--amend`.** The write above records `commitHash: null`. Commit 1 records the findings document and the state together (Git Commit Protocol, `references/shared-conventions.md`). Then record the provenance of that commit in Commit 2, passing the **full 40-character** hash of Commit 1 — an abbreviation is refused rather than expanded, and this call touches nothing but `commitHash`. Add `--epic "{epic}"` when the feature is an epic member — required, per the Pipeline State Protocol.
 
 ```bash
-R="$(bash -c 'for d in "${CLAUDE_PLUGIN_ROOT:-}" "$HOME"/.claude/skills/feature-forge "$HOME"/.claude/plugins/cache/*/feature-forge/* "$HOME"/.claude/plugins/*/feature-forge "$HOME"/.agents/skills/feature-forge ./.agents/skills/feature-forge; do [ -x "$d/scripts/forge-root.sh" ] && exec "$d/scripts/forge-root.sh"; done')"
+R="$(bash -c '[ -z "${FEATURE_FORGE_ROOT:-}" ] || [ -x "$FEATURE_FORGE_ROOT/scripts/forge-root.sh" ] || { echo "feature-forge: FEATURE_FORGE_ROOT=$FEATURE_FORGE_ROOT has no scripts/forge-root.sh" >&2; exit 2; }; for d in "${FEATURE_FORGE_ROOT:-}" "${CLAUDE_PLUGIN_ROOT:-}" "$HOME"/.claude/skills/feature-forge "$HOME"/.claude/plugins/cache/*/feature-forge/* "$HOME"/.claude/plugins/*/feature-forge "$HOME"/.agents/skills/feature-forge ./.agents/skills/feature-forge; do [ -x "$d/scripts/forge-root.sh" ] && exec "$d/scripts/forge-root.sh"; done')"
 [ -n "$R" ] || { echo "feature-forge: cannot locate plugin root" >&2; exit 1; }
 python3 "$R/scripts/forge-session.py" state-verify \
   --feature "{feature}" --stage "{servedStage}" \
@@ -261,26 +263,23 @@ Epic mode is **epic-scoped**, not per-feature: `--stage forge-0-epic` writes `{s
 
 **Served stage.** Pass `--verify-mode` carrying Step 1's explicit or auto-detected mode (`epic`, `prd`, `tech`, `specs`, `backlog`, `impl`); the script maps it to the served production stage. When the caller already owns a stage and states it, additionally pass that value as `--served-stage` — if the two disagree the script fails closed rather than guessing. Derive the served stage **only** from that mode argument or from authoritative pipeline state: conversational context and `currentStage` are never valid inference sources.
 
-**Capability.** Pass `--verify-capability interactive` only when **both** a question mechanism equivalent to `AskUserQuestion` is available **and** a clean-room `forge-verifier` may actually be dispatched right now. Clause (b) tests **permitted dispatch, not a listed tool**: a session that may dispatch only when the user asked, but does have a question mechanism, is `interactive` — the gate's affirmative choice supplies the request. Reserve `manual` for **no** question mechanism **and** **no** permitted dispatch. An auto-verify directive under a no-unsolicited-dispatch bar is presented through the gate and dispatched on the affirmative choice — never skipped, and never resolved by advancing to the production successor.
-
-**Outcome.** Invoke the exit **exactly once**, with the `--outcome` this run's result maps to:
+**Capability.** Determine `--verify-capability` per **Host and capability determination** in `references/stage-exit-protocol.md` (summarized under **Verify Capability** in `references/shared-conventions.md`): pass `interactive` only when a question mechanism equivalent to `AskUserQuestion` is available **and** dispatching a clean-room `forge-verifier` is *permitted* right now — otherwise `manual`. That reference owns the full rule, including why a consent-only bar is still `interactive`. Applied here: an auto-verify directive under a no-unsolicited-dispatch bar is presented through the gate and dispatched on the affirmative choice — never skipped, and never resolved by advancing to the production successor.
 
-| This run's result | `--outcome` |
-|---|---|
-| Zero findings — the artifacts are clean | `passed` |
-| Advisory-only report (no `error`/`gap`) — recorded `passed` with the report attached | `passed` |
-| A report with at least one blocking finding (`error`/`gap`) was written | `findings` |
-| The user explicitly chose to defer pipeline action **and** that skip was persisted via `state-verify --status skipped` | `skipped` |
-| A dispatch, a check, or a state write failed and needs intervention | `failed` |
+**Outcome.** Derive the `--outcome` from the settled on-disk record, never by eyeballing conversational state. **For a production-stage verify** (mode `prd`/`tech`/`specs`/`backlog`/`impl`, so `{servedStage}` is `forge-1-prd`…`forge-5-loop`), after Step 6's `state-verify` write run `select-outcome` — it reads the verify entry and the latest report and returns `{outcome, reason, evidence}`, the `outcome` being exactly one of `passed`, `findings`, `skipped`, or `failed`. Add `--op-failure` when a dispatch, check, or state write failed this run (a `CLEAN_ROOM_UNAVAILABLE` sentinel or a non-answer counts): it is the one signal disk cannot record. On exit 2 the disk names no terminal result and no signal was given — surface its `Error:` line rather than guessing. **Epic mode is out of the verb's scope** — an epic's verification lives in `.epic-state.json`, not a member's `.pipeline-state.json`, so the verb rejects `forge-0-epic`; for `epic` choose the `--outcome` directly per the `### verify` mapping. That four-outcome mapping and its rationale live in `references/select-outcome.md`.
 
-Merely **presenting** blocking findings is `findings`, not `skipped` — all three Step 5 options are `findings`, including (a) "review the findings first", which defers *your* next action rather than the pipeline's. `skipped` is available only for an explicit user deferral of **pipeline action** whose skip has already been persisted. A `CLEAN_ROOM_UNAVAILABLE` sentinel or a non-answer from an advertised dispatch is an operational failure, not a user skip: it is `failed`. A state-write failure is `failed`, and no success block may be printed after one.
+```bash
+R="$(bash -c '[ -z "${FEATURE_FORGE_ROOT:-}" ] || [ -x "$FEATURE_FORGE_ROOT/scripts/forge-root.sh" ] || { echo "feature-forge: FEATURE_FORGE_ROOT=$FEATURE_FORGE_ROOT has no scripts/forge-root.sh" >&2; exit 2; }; for d in "${FEATURE_FORGE_ROOT:-}" "${CLAUDE_PLUGIN_ROOT:-}" "$HOME"/.claude/skills/feature-forge "$HOME"/.claude/plugins/cache/*/feature-forge/* "$HOME"/.claude/plugins/*/feature-forge "$HOME"/.agents/skills/feature-forge ./.agents/skills/feature-forge; do [ -x "$d/scripts/forge-root.sh" ] && exec "$d/scripts/forge-root.sh"; done')"
+[ -n "$R" ] || { echo "feature-forge: cannot locate plugin root" >&2; exit 1; }
+# production-stage modes only (not epic); append --op-failure on an operational failure per the paragraph above
+python3 "$R/scripts/forge-session.py" select-outcome --feature "{feature}" --served-stage "{servedStage}" --skill verify --specs-dir "{specsDir}" --json
+```
 
-Add `--epic "{epic}"` when the feature is an epic member, and `--served-stage "{servedStage}"` when the caller supplied one. Pass no other flags.
+Pass the returned `outcome` straight through as the `stage-exit --outcome` below. Add `--epic "{epic}"` (both calls) when the feature is an epic member, and `--served-stage "{servedStage}"` to the exit when the caller supplied one. Pass no other flags.
 
 **Close this stage with the Scripted Stage Exit** (contract: `references/stage-exit-protocol.md`; do not improvise a "Next steps" list). Run:
 
 ```bash
-R="$(bash -c 'for d in "${CLAUDE_PLUGIN_ROOT:-}" "$HOME"/.claude/skills/feature-forge "$HOME"/.claude/plugins/cache/*/feature-forge/* "$HOME"/.claude/plugins/*/feature-forge "$HOME"/.agents/skills/feature-forge ./.agents/skills/feature-forge; do [ -x "$d/scripts/forge-root.sh" ] && exec "$d/scripts/forge-root.sh"; done')"
+R="$(bash -c '[ -z "${FEATURE_FORGE_ROOT:-}" ] || [ -x "$FEATURE_FORGE_ROOT/scripts/forge-root.sh" ] || { echo "feature-forge: FEATURE_FORGE_ROOT=$FEATURE_FORGE_ROOT has no scripts/forge-root.sh" >&2; exit 2; }; for d in "${FEATURE_FORGE_ROOT:-}" "${CLAUDE_PLUGIN_ROOT:-}" "$HOME"/.claude/skills/feature-forge "$HOME"/.claude/plugins/cache/*/feature-forge/* "$HOME"/.claude/plugins/*/feature-forge "$HOME"/.agents/skills/feature-forge ./.agents/skills/feature-forge; do [ -x "$d/scripts/forge-root.sh" ] && exec "$d/scripts/forge-root.sh"; done')"
 [ -n "$R" ] || { echo "feature-forge: cannot locate plugin root" >&2; exit 1; }
 python3 "$R/scripts/forge-session.py" stage-exit --feature "{feature}" --stage forge-verify --owner "{owner}" --outcome "{VerifyOutcome}" --verify-mode "{mode}" --specs-dir "{specsDir}" --host claude --verify-capability "{verify-capability}"
 ```
@@ -298,7 +297,7 @@ Obey the DIRECTIVES it prints, in the consumption order this protocol fixes: sur
 - For specs verification, also run the deterministic traceability validator to supplement agent-driven traceability checks. Include any uncovered requirements or orphaned references as findings:
 
 ```bash
-R="$(bash -c 'for d in "${CLAUDE_PLUGIN_ROOT:-}" "$HOME"/.claude/skills/feature-forge "$HOME"/.claude/plugins/cache/*/feature-forge/* "$HOME"/.claude/plugins/*/feature-forge "$HOME"/.agents/skills/feature-forge ./.agents/skills/feature-forge; do [ -x "$d/scripts/forge-root.sh" ] && exec "$d/scripts/forge-root.sh"; done')"
+R="$(bash -c '[ -z "${FEATURE_FORGE_ROOT:-}" ] || [ -x "$FEATURE_FORGE_ROOT/scripts/forge-root.sh" ] || { echo "feature-forge: FEATURE_FORGE_ROOT=$FEATURE_FORGE_ROOT has no scripts/forge-root.sh" >&2; exit 2; }; for d in "${FEATURE_FORGE_ROOT:-}" "${CLAUDE_PLUGIN_ROOT:-}" "$HOME"/.claude/skills/feature-forge "$HOME"/.claude/plugins/cache/*/feature-forge/* "$HOME"/.claude/plugins/*/feature-forge "$HOME"/.agents/skills/feature-forge ./.agents/skills/feature-forge; do [ -x "$d/scripts/forge-root.sh" ] && exec "$d/scripts/forge-root.sh"; done')"
 [ -n "$R" ] || { echo "feature-forge: cannot locate plugin root" >&2; exit 1; }
 python3 "$R/scripts/validate-traceability.py" {resolvedFeatureDir}/PRD.md {resolvedFeatureDir}/ --json
 ```

package/adapters/claude/skills/forge-verify/references/findings-template.md

@@ -4,7 +4,7 @@ Loaded by the **parent orchestrator** role of `forge-verify` at Step 4 (write th
 
 ## Truncated Verifier Returns (Synthesize gate)
 
-The Agent tool hands the parent **only the verifier's final message**. A verifier that
+The `Agent` tool hands the parent **only the verifier's final message**. A verifier that
 ends its run on a status line instead of the report returns *that line* as the entire
 result — the digest it built exists in its transcript but never reaches you. Observed
 in the wild (issue #183): two ~130k-token, 40+-tool-call runs each returned a single
@@ -131,10 +131,10 @@ violate REQ-STATE-02; per-feature status is always derived live from each member
 `.pipeline-state.json`).
 
 Set `stages.forge-verify-epic.status` to `findings-reported` when the report lists at
-least one **blocking** finding (`error`/`gap`), else `passed` — for an advisory-only
-report pass `--findings-file`/`--findings-count` alongside `--status passed`, exactly
-as in feature mode (the severity floor in `skills/forge-verify/SKILL.md`) — recording
-`findingsFile`, `findingsCount`, and `verifiedAt`.
+least one **blocking** finding (`error`/`gap`), else `passed`. Always attach the report
+and its total count, including count `0` for a clean report, exactly as in feature mode
+(the severity floor in `skills/forge-verify/SKILL.md`) — recording `findingsFile`,
+`findingsCount`, and `verifiedAt`.
 
 **Write it with `state-verify`, never by hand.** `--stage forge-0-epic` is the sanctioned
 epic writer: it creates the file lazily, mutates only `stages.forge-verify-epic` plus the
@@ -145,12 +145,12 @@ This is the one `state-*` call site where the member `--epic` rule does **not**
 version — which is what keeps the result reading fresh rather than stale.
 
 ```bash
-R="$(bash -c 'for d in "${CLAUDE_PLUGIN_ROOT:-}" "$HOME"/.claude/skills/feature-forge "$HOME"/.claude/plugins/cache/*/feature-forge/* "$HOME"/.claude/plugins/*/feature-forge "$HOME"/.agents/skills/feature-forge ./.agents/skills/feature-forge; do [ -x "$d/scripts/forge-root.sh" ] && exec "$d/scripts/forge-root.sh"; done')"
+R="$(bash -c '[ -z "${FEATURE_FORGE_ROOT:-}" ] || [ -x "$FEATURE_FORGE_ROOT/scripts/forge-root.sh" ] || { echo "feature-forge: FEATURE_FORGE_ROOT=$FEATURE_FORGE_ROOT has no scripts/forge-root.sh" >&2; exit 2; }; for d in "${FEATURE_FORGE_ROOT:-}" "${CLAUDE_PLUGIN_ROOT:-}" "$HOME"/.claude/skills/feature-forge "$HOME"/.claude/plugins/cache/*/feature-forge/* "$HOME"/.claude/plugins/*/feature-forge "$HOME"/.agents/skills/feature-forge ./.agents/skills/feature-forge; do [ -x "$d/scripts/forge-root.sh" ] && exec "$d/scripts/forge-root.sh"; done')"
 [ -n "$R" ] || { echo "feature-forge: cannot locate plugin root" >&2; exit 1; }
 python3 "$R/scripts/forge-session.py" state-verify \
   --feature "{epic}" --stage forge-0-epic \
-  --status "{passed|findings-reported}" \
-  --findings-file "{relative findings path}" --findings-count {n} \
+  --status "{outcome-specific status}" \
+  --findings-file "{relative findings path}" --findings-count {outcome-specific count} \
   --verified-stage-version {manifest revision} --specs-dir "{specsDir}"
 ```
 

package/adapters/claude/skills/forge-verify/references/select-outcome.md

@@ -0,0 +1,138 @@
+# `select-outcome` — deterministic exit-outcome selection
+
+`forge-verify` and `forge-fix` each close a run by choosing exactly one
+`stage-exit --outcome`. That choice used to be model judgment over conversational
+state. `forge-session.py select-outcome` derives it from the authoritative on-disk
+record instead — the same "never eyeball the graph" move `rank-features` makes when
+it derives `verifyGate` for the navigator (issue #276, part of the #265 guard-first
+program).
+
+The verb reads, never writes. It is called at the exit step, **after** the run's
+result has already been persisted by `state-verify` and the findings document has
+been written, so the disk state it reads is the run's own settled record.
+
+## Synopsis
+
+```
+python3 forge-session.py select-outcome \
+  --feature F --served-stage S --skill verify|fix \
+  [--op-failure] [--user-deferred] [--decisions-open] \
+  [--specs-dir DIR] [--epic E] [--json]
+```
+
+`--served-stage` is one of the five verify-token production stages
+(`forge-1-prd`, `forge-2-tech`, `forge-3-specs`, `forge-4-backlog`, `forge-5-loop`).
+`forge-0-epic` (epic-scoped verification lives in `.epic-state.json`) and
+`forge-6-docs` (no verify token) are **out of scope** and rejected.
+
+Output — always `{outcome, reason, evidence: [...]}`:
+
+```json
+{
+  "outcome": "reverify-findings",
+  "reason": "fixes were applied and a re-verify reopened blocking findings for the served stage",
+  "evidence": [
+    "forge-verify-tech status = findings-reported",
+    "latest report .verification/VERIFY-tech-2026-09-07-round2.md (round 2): 2 blocking = 1 error + 1 gap, 3 total finding(s)",
+    "round reports on disk: 2",
+    "fix progress: 1 step(s) applied; sweep 1 fixed / 0 justified / 0 false-positive"
+  ]
+}
+```
+
+## Outcome vocabularies
+
+The verb chooses from exactly the branch skills' `--outcome` domains — the
+`VerifyOutcome` and `FixOutcome` aliases in `forge-session.py`. Those aliases are
+pinned to the tables below by `tests/test_select_outcome.py` (both directions), so the
+enum and this reader-facing mapping can never drift. `forge-verify` and `forge-fix`
+used to carry these tables in their SKILL bodies; issue #278 flipped the bodies to
+call the verb and moved the tables here, their reader-facing home.
+
+### verify
+
+| This run's result | `--outcome` |
+|---|---|
+| Zero findings — the artifacts are clean | `passed` |
+| Advisory-only report (no `error`/`gap`), recorded `passed` with the report attached | `passed` |
+| A report with at least one blocking finding (`error`/`gap`) was written | `findings` |
+| The user explicitly deferred pipeline action and that skip was persisted (`state-verify --status skipped`) | `skipped` |
+| A dispatch, check, or state write failed and needs intervention | `failed` |
+
+### fix
+
+| This run's result | `--outcome` | Authoritative downstream action |
+|---|---|---|
+| No applicable findings document or steps (Step 1.5) | `no-findings` | Re-verify while verification is still owed; otherwise the live production successor |
+| User decisions remain unresolved (Step 3) | `decisions` | Resume `forge-fix` naming the unresolved decisions; no advancement |
+| A fix step, validation, commit, or state write failed (Steps 4–6) | `failed` | Fix/navigator recovery; no advancement |
+| Fixes persisted and re-verify not yet run (nested, or manual capability) | `applied` | Re-run `forge-verify` for the same served stage — re-verification is mandatory |
+| A mandatory re-verify passed | `reverified` | The live production successor |
+| A mandatory re-verify reopened an unresolved prior finding or a new blocking defect (clean-or-advisory-only is `reverified`) | `reverify-findings` | `forge-fix` for the same served stage |
+| The user explicitly deferred the fix pass or the re-verify (Steps 3, 6) | `deferred` | Deterministic `forge-fix`/navigator resume; findings remain unresolved; no advancement |
+
+## Inputs read from disk
+
+- The served stage's `stages.forge-verify-{token}` entry in `.pipeline-state.json` —
+  its `status` is the authoritative signal, written mechanically by `state-verify`.
+- The latest `.verification/VERIFY-{mode}-*.md` report (round-aware: `-round{N}` beats
+  the un-suffixed base file, and the newest day's newest round is "latest"). Blocking
+  counts come from the `## Findings` severities (`error`/`gap`), never from the human
+  `## Summary` line.
+- Each report's `## Fix Progress` section — its `[APPLIED]` steps and the sweep
+  dispositions (`FIXED`/`JUSTIFIED`/`FALSE-POSITIVE`). A report carrying `[APPLIED]`
+  steps is the disk record that a fix pass ran (`fix-applied`).
+
+## Derivation
+
+### verify
+
+1. `--op-failure` → **failed**.
+2. entry `skipped` → **skipped**.
+3. entry `findings-reported` → **findings**.
+4. entry `passed` → **passed** (a clean report, an advisory-only report, or accepted
+   residual findings — all three are recorded `passed`).
+5. otherwise (no terminal result recorded) → **fail closed** (exit 2).
+
+### fix
+
+1. `--op-failure` → **failed**.
+2. `--user-deferred` → **deferred**.
+3. `--decisions-open` → **decisions**.
+4. no report for the mode → **no-findings**.
+5. entry `findings-applied` → **applied** (a re-verify is still owed).
+6. entry `passed` → **reverified**. A forge-fix pass reaches a `passed` served stage
+   only by re-verifying it: forge-fix Step 1.5 resolves its served stage to an
+   *unresolved* verify entry (`findings-reported`, `findings-applied`, or a pending
+   debt), never an already-resolved `passed`, so a `passed` entry at a fix exit is the
+   mandatory re-verify's own passing result. (`fix-applied` is not consulted here — it
+   reads every round on disk and so cannot tell this pass's fixes from an earlier
+   cycle's.)
+7. entry `findings-reported` + a fix ran → **reverify-findings** (a re-verify reopened
+   findings); entry `findings-reported` + no fix ran → **fail closed** (exit 2).
+8. any other resolved/absent entry with no fix → **no-findings**.
+
+Two cases the issue calls out because a model tends to mis-eyeball them, and the verb
+gets right by keying on the entry status rather than the presence of findings text:
+
+- **A sweep whose survivors are all `FALSE-POSITIVE`.** Disposition is not a defect —
+  the entry is still `findings-applied`, so the outcome is **applied**, never `failed`.
+- **A re-verify red on a scoped subset.** The round-2 report reopens blocking findings
+  and the entry returns to `findings-reported`, so the outcome is **reverify-findings**.
+
+## Runtime signals
+
+Three outcomes cannot be read off disk, because the fact that distinguishes each is
+inherently runtime:
+
+- **failed** — an operational failure (a dispatch, step, validation, commit, or state
+  write) leaves *no* state write, so disk shows the prior state.
+- **deferred** — an explicit user deferral of the fix or re-verify is byte-identical on
+  disk to a nested/manual `applied` (both leave the entry at `findings-applied`).
+- **decisions** — unresolved user decisions live in the free-prose "User Decisions
+  Required" section, which has no machine format.
+
+The caller asserts these with `--op-failure` / `--user-deferred` / `--decisions-open`
+(the latter two are `fix`-only). Everything else is pure disk derivation. When the disk
+record names no terminal outcome and no signal was supplied, the verb **fails closed**
+with `Error:` on stderr and exit 2 — it never guesses.