@garygentry/feature-forge 0.3.2 → 0.3.4

package/adapters/claude/skills/forge-fix/references/stage-exit-protocol.md

@@ -47,8 +47,8 @@ an epic member. Only the flags below are stage-specific; pass no others.
 |---|---|---|
 | `forge-0-epic` | `forge-0-epic` | `--next-feature "{member}"` when a concrete member exists |
 | `forge-1-prd` … `forge-4-backlog` | that stage's own id | none beyond identity/capability |
-| `forge-5-loop` | `forge-5-loop` | `--outcome` — one of `complete`, `partial`, `blocked`, `needs-human`, `deferred` |
-| `forge-6-docs` | `forge-6-docs` | `--outcome` — `complete` or `blocked` |
+| `forge-5-loop` | `forge-5-loop` | `--outcome` — one of `complete`, `partial`, `blocked`, `needs-human`, `deferred`, `resolved`; optional `--cause dependency-starvation` with `--outcome partial` |
+| `forge-6-docs` | `forge-6-docs` | `--outcome` — `complete`, `blocked`, or `skipped` (deliberate docs skip, persisted via `state-skip` before the exit; routes like `complete` with honest wording) |
 | direct `forge-verify` | `forge-verify` | `--owner direct`, `--outcome` (`passed`, `findings`, `skipped`, `failed`), and served-stage metadata |
 | nested `forge-verify` | `forge-verify` | `--owner nested`, plus the same outcome and served-stage metadata |
 | direct/nested `forge-fix` | `forge-fix` | the matching `--owner`, a `FixOutcome` (`no-findings`, `decisions`, `failed`, `applied`, `reverified`, `reverify-findings`, `deferred`), and served-stage metadata |
@@ -100,9 +100,9 @@ Obey the DIRECTIVES it prints, in the consumption order this protocol fixes: sur
 
 The stamp is shown with `--host claude`; the adapter build substitutes `pi`/`generic` per
 target, and §"Host and capability determination" below governs the value. The literal is
-deliberate — `scripts/build-adapters.py` keys its host translation on the exact string
-`--host claude`, and the stamp sites are compared byte-for-byte, so it is the one token in
-that line that is not a placeholder.
+deliberate — `scripts/build-adapters.py` keys its host translation on the exact canon
+value of that flag, and the stamp sites are compared byte-for-byte, so it is the one token
+in that line that is not a placeholder.
 
 ## Host and capability determination
 
@@ -110,8 +110,9 @@ Before the call, compute the two inputs independently. They are unrelated: **a h
 implies a capability**, and the script takes `--verify-capability` at face value.
 
 **`--host`** describes only the active adapter command surface — `claude`, `pi`, or
-`generic`. It selects command syntax (`/feature-forge:` vs `/skill:` vs host-neutral) and
-fresh-session wording (`/clear` vs `/new` vs neutral prose). Nothing else.
+`generic`. It selects command syntax (Claude's stage-command prefix vs Pi's `/skill:` vs
+host-neutral) and fresh-session wording (Claude's clear command vs Pi's `/new` vs neutral
+prose). Nothing else.
 
 **`--verify-capability interactive`** is passed only when **both** of these hold:
 
@@ -151,7 +152,8 @@ production successor** while verification is unresolved.
 - a capable Pi session is `--host pi --verify-capability interactive`, and receives the
   same logical gate a capable Claude session does;
 - Pi without a dispatchable verifier is `--host pi --verify-capability manual`;
-- a Claude session that cannot dispatch is `--host claude --verify-capability manual`.
+- a Claude session that cannot dispatch keeps the Claude host value with
+  `--verify-capability manual`.
 
 Interactive gate options keep their explicit labels, their recommended default, and their
 one-line trade-off descriptions (below). The manual path prints the verify command as the
@@ -199,6 +201,44 @@ second sentinel-terminated block **inside** an outer stage's exit, breaking the
 exactly-one-terminal-block rule — and the canon guard cannot catch it, because both
 wordings legitimately appear in the same file. Judge the token, not the phrasing.
 
+## Caller-side resumption: the declared resume point
+
+The `owner:` token and `terminalOwnedBy` arbitrate who prints — but they specify only
+the **callee** side: a nested skill stays quiet and returns its structured result. This
+section is the reciprocal, caller-side half of that contract.
+
+**On a sub-skill's return, the caller re-owns the terminal.** Every closing instruction
+in the callee's own body — its report-and-stop posture, its "confirm the result" close,
+its own next-steps habits — is void for this turn. The caller resumes at its **declared
+resume point**, in the same turn, and its remaining steps run to its own terminal.
+
+**Every Skill-tool delegation site declares, at the invocation, what happens on
+return.** Suppression without resumption is the failure mode this section closes: the
+callee's closing posture is the freshest instruction in context while the caller's next
+step is the oldest, so an undeclared return silently ends the run one layer too early —
+the caller's remaining steps (validation, state writes, commit, stage exit) dropped,
+with no error surfaced. An implicit "let it run to its natural stopping point" is that
+bug spelled politely, and it is banned on delegation sites. A site takes one of two
+declared postures:
+
+- **Delegate-and-resume** — the callee is a sub-step of the caller: the site names the
+  caller's own step that control returns to, and the caller continues there in the same
+  turn. The callee never owns the caller's terminal. The worked instance is
+  `forge-4-backlog` Step 4's **Return contract** for the `author-backlog` delegation
+  (control returns at Step 5; the sub-skill's direct-invocation posture — its approval
+  gate and its validate-and-confirm close — is explicitly disapplied on the delegated
+  path). New delegation sites follow that pattern rather than re-deriving it.
+- **Terminal handoff** — the caller's job ends at the invocation and the invoked skill
+  owns the terminal from there on (e.g. the navigator's `autoInvokeNextStage`
+  "continue in this session" advance into the next production stage). Declaring the
+  handoff is what keeps it distinct from an accidental drop.
+
+Scope: this contract governs **Skill-tool delegation within one session**. The
+truncated-verifier-return guard (forge-verify's `findings-template.md`, "Truncated
+Verifier Returns") is a different mechanism — it polices what an **Agent-dispatched
+subagent's** return payload must contain, not where a caller resumes. A dispatch site
+can be subject to both; satisfy each on its own terms.
+
 ## Directive consumption order
 
 `stage-exit` emits a DIRECTIVES object and (for a direct owner) a NEXT-STEPS block. The
@@ -299,7 +339,11 @@ first:
   a stage's artifact commit.
 - **Skip for now** — go straight to the NEXT-STEPS block without verifying. Record this
   stage's verify status as `skipped` in pipeline state (via `state-verify`, never by hand)
-  **only** on an explicit skip — a skip does not go stale.
+  **only** on an explicit skip — a skip does not go stale. Exception: if the existing
+  entry records `passed` or `findings-applied` (a resolved result whose freshness has
+  merely lapsed), write **nothing** — `state-verify` refuses to demote a resolved status
+  to `skipped` (#203), and the recorded result stands on its own; the user's decline is
+  honored by simply not re-verifying.
 
 **Advancement is allowed only after a pass, or after an explicit skip has been
 persisted.** Choosing to stop, or losing the interaction, produces no advancing terminal
@@ -358,8 +402,8 @@ directive is informational — you do **not** re-derive the wording:
   unchanged; the block appends a non-blocking reminder line ("You also flagged N epic
   change(s) to reconcile when convenient …"). This is *finish-then-edit*.
 
-Either way the added lines are host-neutral (no literal `/clear`) and sit **above** the
-sentinel; just print the NEXT-STEPS block verbatim as always.
+Either way the added lines are host-neutral (they name no fresh-session command) and sit
+**above** the sentinel; just print the NEXT-STEPS block verbatim as always.
 
 ### Deferred decisions — do not solicit next-stage decisions at this exit
 

package/adapters/claude/skills/forge-guide/references/forge-config-schema.json

@@ -67,6 +67,12 @@
       "default": true,
       "description": "When true (default), the /feature-forge:forge navigator auto-invokes the next pipeline stage via the Skill tool after the user confirms it, instead of only printing the command to copy. Set false to keep the old copy-paste behavior (the navigator suggests the command but never launches it). Ignored on non-Claude hosts, which always fall back to printing the command."
     },
+    "docsStage": {
+      "type": "string",
+      "enum": ["prompt", "skip"],
+      "default": "prompt",
+      "description": "How forge-6-docs' Documentation Decision Gate behaves at stage entry. \"prompt\" (default) asks generate-vs-skip once per run; \"skip\" records the stage as deliberately skipped with no prompt — the skill runs the state-skip verb and closes the pipeline honestly (stage-exit --outcome skipped), so docs are never implied-mandatory for teams that document differently. The gate's refuse-to-erase guard still applies: docs that already exist are never silently reclassified, whatever this is set to. Read directly by forge-6-docs (not part of loopRunner resolution); an unrecognized or absent value behaves as \"prompt\"."
+    },
     "autoVerify": {
       "type": "boolean",
       "default": false,
@@ -195,6 +201,18 @@
           "default": "",
           "description": "Project-default coding agent id, so a project can fix its agent once without specifying it every run. Empty string ⇒ no project default (the runner's own default — claude-cli for rauf — applies, behaving exactly as today). Overridden by the per-run agent selector (run > project precedence, resolved inside forge before the single --agent is emitted). Ignored when agentArgument is absent."
         },
+        "agentMode": {
+          "type": "string",
+          "enum": ["prompt", "auto"],
+          "default": "prompt",
+          "description": "Gates forge-5's Step 2d agent AskUserQuestion. \"prompt\" (default) presents the agent question — byte-identical to today. \"auto\" suppresses ONLY the interactive pick: the agent resolves from defaultAgent (or the runner default when unset), and the availability probe, the verdict classification (UNKNOWN hard-reject before side-effects / UNAVAILABLE proceed-or-choose / probe-failure handling), and the Claude-alias model guard ALL still run; the resolved 'Agent: {id} (source: …)' confirmation line still shows, so the choice is never hidden. Meaningless when agentArgument is absent/empty — that capability gate already removes the entire agent surface, and this field adds no second gate. An unrecognized value behaves as \"prompt\"."
+        },
+        "reviewMode": {
+          "type": "string",
+          "enum": ["prompt", "always", "never"],
+          "default": "prompt",
+          "description": "Gates forge-5's Step 2d Run-mode AskUserQuestion (rauf only — ignored when name != \"rauf\", where no Run-mode question exists). \"prompt\" (default) surfaces the question with the review pass pre-selected — byte-identical to today. \"always\" skips the question and appends --review unconditionally; the rendered command in the confirmation still shows --review, so the choice is never hidden. \"never\" skips the question and launches the bare rendered command. Under \"always\"/\"never\", a NARROWER situational retry-blocked question still surfaces when — and only when — the backlog has blocked items (the --review decision itself is fixed by the mode and never re-asked); with no blocked items the launch is prompt-free. An unrecognized value behaves as \"prompt\"."
+        },
         "preconditionFile": {
           "type": "string",
           "default": ".rauf.json",

package/adapters/claude/skills/forge-guide/references/ralph-loop-contract.md

@@ -58,9 +58,12 @@ defined authoritatively in rauf's
 > item aside and **keeps working other runnable items to completion** (rauf:
 > `runner.ts` needs_human handler). So a supervising session can surface those
 > events live (visibility) and cancel early, but it cannot inject an answer and
-> resume the set-aside item mid-run — resolution is a follow-up retry pass. A
-> first-class pause/resume-with-answer capability is a desirable runner
-> enhancement (see `plans/rauf-enhancement-recommendations.md`).
+> resume the set-aside item mid-run — resolution is the **Post-Run Recovery
+> Procedure** (`skills/forge-5-loop/references/recovery-procedure.md`): record the
+> answer via `decision-record` at the moment of collection, then drive recovery
+> from the record after the run ends. A first-class pause/resume-with-answer
+> capability is a desirable runner enhancement (see
+> `plans/rauf-enhancement-recommendations.md`).
 
 ## rauf is the default and reference implementation
 

package/adapters/claude/skills/forge-guide/references/shared-conventions.md

@@ -36,7 +36,7 @@ I found that the codebase uses React and TanStack Router.
 
 ### Decision Support: Help the User Choose
 
-When an `AskUserQuestion` carries substantive options (a real choice — not a trivial yes/no confirmation), do not just list them. The interview stages have already done codebase research and integration analysis; surfacing that synthesis at the decision moment is the whole point. For every such question:
+When a question posed through `AskUserQuestion` carries substantive options (a real choice — not a trivial yes/no confirmation), do not just list them. The interview stages have already done codebase research and integration analysis; surfacing that synthesis at the decision moment is the whole point. For every such question:
 
 - **Lead with a recommended option.** Place it first and label it `(recommended)` (matching the `AskUserQuestion` "(Recommended)" convention).
 - **Put the trade-off in each option's `description`.** Say why you'd pick it and what you give up versus the alternatives — the cost, not just the benefit.
@@ -53,6 +53,17 @@ For genuinely comparable artifacts (competing module structures, two code snippe
 
 The **Branch Setup** block below is the reference pattern: a strong recommendation as the first option, rationale inline, the alternative still available, never a hard-stop.
 
+## Stage Review Gate
+
+Every authoring stage ends its "Review with User" step in exactly one of two shapes. Which shape a stage uses is declared **here, once** — each stage's review step points at this block by title, and the shape is never re-derived from the surrounding prose or inferred from how sibling stages behave (three stages block and one does not; majority-shape inference is precisely the failure this block exists to prevent):
+
+- **Blocking review (gate).** The stage presents the artifact and collects feedback through `AskUserQuestion`; it does **not** proceed until the user answers, iterating until they confirm. Stages: **forge-1-prd** (Step 5), **forge-2-tech** (Step 6), **forge-3-specs** (Step 6).
+- **Non-blocking review (invitation).** The stage states the artifact is ready and invites adjustments **as a statement, not a question** — and then **proceeds to the next step in the same turn unless the user asks for changes**. The invitation obliges the agent to *continue*: emitting the invitation sentence and stopping treats the non-gate as a gate and strands the stage `in-progress` with its completion step unrun — a defect, not caution. Stage: **forge-4-backlog** (Step 6).
+
+**Why the shapes differ.** A blocking review guards an artifact whose content was just authored from open-ended interview or synthesis — the user is the only authority on "complete", so the stage must wait. forge-4's backlog is *derived* from specs the user already approved, was planned interactively in its Step 3, and is machine-validated in its Step 5; a second hard gate would re-ask a settled question, and the loop never launches without forge-5-loop's own Step 2d confirmation anyway. The invitation is a courtesy checkpoint, not an approval gate (removed deliberately in #78's consistency sweep).
+
+A stage that changes shape changes it **in this block first**; the per-stage pointer stays a pointer.
+
 ## Configuration Reading
 
 Read `forge.config.json` from the project root. If it doesn't exist, use defaults.
@@ -130,7 +141,7 @@ mkdir -p "<specsDir>"
 [ -f "<specsDir>/AGENTS.md" ] || cp "$R/references/templates/specs-hygiene/AGENTS.md" "<specsDir>/AGENTS.md"
 ```
 
-If the host is Claude (the `AskUserQuestion` tool is available), also ensure the Claude-framed variant:
+If the host is Claude (the Claude-native question tool is available), also ensure the Claude-framed variant:
 
 ```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')"
@@ -191,7 +202,7 @@ Pipeline state is written by the `state-*` verbs of `scripts/forge-session.py`
 
 If a `state-*` verb exits 2, surface the plain `Error:` line from stderr verbatim, do **not** proceed to the next step of the surrounding protocol, and do **not** hand-author the JSON as a workaround. The stage remains resumable because the entry stamp is already on disk — re-run the verb once the cause is fixed.
 
-**The eight `state-*` verbs.** `state-enter` (Stage-Entry Guard), `state-artifact` (incremental artifact tracking), `state-complete` (Git Commit Protocol), `state-branch` (Branch Setup and Branch Reconciliation), `state-note` (the Immediate Downstream Note below, and the optional completion note at stage closure), `state-decision` (deferred decisions), `state-ecr` (epic change requests), and `state-verify` (one `forge-verify-*` verification transition — below). The `--epic` member requirement and the exit-2 failure protocol above apply to **every** one of them, `state-verify` included; no verify entry is ever hand-authored.
+**The nine `state-*` verbs.** `state-enter` (Stage-Entry Guard), `state-artifact` (incremental artifact tracking), `state-complete` (Git Commit Protocol), `state-skip` (the deliberate forge-6-docs documentation skip — scoped to that one stage; writes `status: "skipped"` + `skippedAt`, refuses to erase a record of docs that exist, and is the only sanctioned writer of a skipped docs stage), `state-branch` (Branch Setup and Branch Reconciliation), `state-note` (the Immediate Downstream Note below, and the optional completion note at stage closure), `state-decision` (deferred decisions), `state-ecr` (epic change requests), and `state-verify` (one `forge-verify-*` verification transition — below). The `--epic` member requirement and the exit-2 failure protocol above apply to **every** one of them, `state-verify` included; no verify entry is ever hand-authored.
 
 ### `state-verify` — verification results and provenance
 
@@ -395,7 +406,7 @@ Invoke this block **at the head of any post-entry step that writes a stage artif
 
 1. **Proceed** when `stages.{stage}.status` is `"in-progress"` (this session's Entry Stamp — you are finishing the run you started) or absent/`pending`. Run the write / exit normally.
 
-2. **Detect-and-refuse** when ALL of these hold: `stages.{stage}.status ∈ {"complete", "stale"}` **AND** the stage's artifacts (incl. `TRACEABILITY.md` for forge-3-specs) exist on disk **AND** a `commitHash` is recorded for the stage **AND** you did **not** author this stage earlier in the current session. This is a stale/replayed continuation of an already-finished, committed stage. Do **not** overwrite the artifact or re-run the exit. Route instead to the **Stage-Entry Guard**'s *Re-authoring* path: surface the same `AskUserQuestion` warning ("A completed {stage} artifact already exists for '{feature}' (v{n}{, marked stale}). Continuing will create a new version. Proceed?"). Only on explicit confirmation re-enter from the Entry Stamp (the version bumps at exit); otherwise **stop** and report that the stage is already complete — cite the recorded `commitHash` and offer `/feature-forge:forge {feature}` to see true state.
+2. **Detect-and-refuse** when ALL of these hold: `stages.{stage}.status ∈ {"complete", "stale"}` **AND** the stage's artifacts (incl. `TRACEABILITY.md` for forge-3-specs) exist on disk **AND** a `commitHash` is recorded for the stage **AND** you did **not** author this stage earlier in the current session. This is a stale/replayed continuation of an already-finished, committed stage. Do **not** overwrite the artifact or re-run the exit. Route instead to the **Stage-Entry Guard**'s *Re-authoring* path: surface the same warning via `AskUserQuestion` ("A completed {stage} artifact already exists for '{feature}' (v{n}{, marked stale}). Continuing will create a new version. Proceed?"). Only on explicit confirmation re-enter from the Entry Stamp (the version bumps at exit); otherwise **stop** and report that the stage is already complete — cite the recorded `commitHash` and offer `/feature-forge:forge {feature}` to see true state.
 
 When you cannot confirm you authored the current run, treat it as a replay and refuse: a false refuse costs one confirmation click; a false proceed overwrites a committed artifact and re-churns a stage version. `--force` follows Force Mode (skip the gate, treat as a deliberate re-author).
 

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

@@ -30,7 +30,7 @@ Pick based on how many checks the mode carries (see the per-mode totals in Step
 - **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 38, backlog 27, impl 23): parallel dimensioned fan-out.**
+- **Large modes (specs 39, backlog 29, impl 25): parallel dimensioned fan-out.**
   Split the mode's checklist into **dimension groups** and dispatch **one
   `forge-verifier` per group, in parallel — a single message with multiple Agent
   calls** (the `superpowers:dispatching-parallel-agents` pattern). Each instance owns a
@@ -38,14 +38,14 @@ Pick based on how many checks the mode carries (see the per-mode totals in Step
   run concurrently. Suggested groups (map to the category clusters in that mode's own
   checklist file):
   - **specs** (`references/verification-checklists/specs.md`): (1) types/contracts,
-    (2) architecture/layout, (3) cross-reference &
-    traceability, (4) testing strategy, (5) integration.
+    (2) architecture/layout, (3) cross-reference & traceability (owns CHECK-S39),
+    (4) testing strategy, (5) integration.
   - **backlog** (`references/verification-checklists/backlog.md`): (1) item scoping &
     acceptance criteria, (2) dependency/ordering sanity,
-    (3) spec coverage & traceability, (4) schema/enum correctness.
-  - **impl** (`references/verification-checklists/impl.md`): (1) requirement coverage vs specs, (2) integration correctness,
-    (3) testing, (4) code-quality/conventions, (5) runnability (owns CHECK-I21/I22 —
-    the smoke command and the non-test-caller heuristic).
+    (3) spec coverage & traceability (owns CHECK-B29), (4) schema/enum correctness.
+  - **impl** (`references/verification-checklists/impl.md`): (1) requirement coverage vs specs (owns CHECK-I24/I25),
+    (2) integration correctness, (3) testing, (4) code-quality/conventions,
+    (5) runnability (owns CHECK-I21/I22 — the smoke command and the non-test-caller heuristic).
 
   In each parallel instance's prompt, pass: the feature, the mode, the **dimension
   label**, the **exact CHECK-IDs it owns**, and a note that **it is one of several
@@ -56,8 +56,7 @@ Pick based on how many checks the mode carries (see the per-mode totals in Step
 
 ### Synthesize (parent session)
 
-The verifier(s) are read-only — they return findings as their response; **you** (the
-parent) assemble and write the single document to
+The verifier(s) are read-only — they return findings as their response. **Gate every return first** through "Truncated Verifier Returns" in `references/findings-template.md`: a return without the report structure is a dropped digest (issue #183) — resume the agent via `SendMessage` (or re-dispatch), and never synthesize, write a document, or record a verify result from a truncated return. Then **you** (the parent) assemble and write the single document to
 `{resolvedFeatureDir}/.verification/VERIFY-{mode}-{YYYY-MM-DD}.md`. When you fanned out:
 1. Concatenate all instances' findings and **renumber `V-NNN` IDs uniquely** across the
    merged set.
@@ -169,7 +168,7 @@ Load into context ALL artifacts for this feature based on mode:
 
 Read `references/verification-checklists/{mode}.md` for the detailed checklist for the mode being verified — one of `references/verification-checklists/prd.md`, `references/verification-checklists/tech.md`, `references/verification-checklists/specs.md`, `references/verification-checklists/backlog.md`, `references/verification-checklists/impl.md`, `references/verification-checklists/epic.md`. Read only that mode's file. Execute every check. Do not skip checks because things "look fine." **Exception — a re-verify is scoped, not a fresh sweep:** when the served stage's verify entry is `findings-applied`, follow "Re-verify scope and convergence" in `references/stage-exit-protocol.md` — confirm the prior report's findings against their acceptance evidence and examine the fix's own delta; only an unresolved prior finding or a new blocking defect the fix itself introduced may block, every other observation is advisory, and a finding with a recorded decision is never re-filed. The orchestrator-only **Findings Document Template (Step 4)**, worked **Example Findings (Step 4)**, and **Epic Mode State Write Detail (Step 6)** sections live in `references/findings-template.md`, read later by the parent role at Steps 4/6.
 
-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: 38 checks, backlog: 27 checks, impl: 23 checks, epic: 10 checks), you likely skipped checks — go back and complete them.
+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.
 
@@ -184,7 +183,7 @@ Every finding must include:
 
    **Severity floor (anti-churn).** An inaccuracy confined to comments, docstrings, or test narration — prose no runtime path executes and no decision consumes — caps at `inconsistency`, never `error`. It is worth recording, but a wrong sentence beside correct code does not block a stage the way wrong behavior does. **Routing consequence:** `error` and `gap` are the two **blocking** severities; `inconsistency` and `improvement` are **advisory**. A report with at least one blocking finding records `findings-reported` and routes to forge-fix. A report whose findings are all advisory records `passed` **with the report still attached** (Step 6) and the pipeline advances — an advisory-only report never fences a fix round. A **meta-guard** (a test protecting other tests or prose) is judged against its declared protection set: guard-incompleteness against a declared non-goal is never a finding (`references/stage-exit-protocol.md` § Re-verify scope).
 
-   **A checklist item with no PRD position behind it is a PRD gap, not a design to invent.** Several checks are deliberately conditional — `CHECK-S27` ("Concurrent access scenarios are addressed **if relevant**") is the clearest, and the same shape appears for performance, observability, and security checks that defer to the PRD. When such a check fires and the PRD takes **no position** on the concern, the finding is that the *requirements* are silent. Report it as an `improvement` (or a `gap` against the PRD, in `prd` mode) whose suggested fix is to **record the position** — including "out of scope, single writer assumed", which is a complete answer. Do **not** specify a mechanism to satisfy the check: a verifier that answers an open requirements question by designing a protocol converts a one-sentence PRD amendment into a foundational change that no requirement asked for, at the stage where it is least visible. Precedent both ways: `epic-orchestration` V-008 raised `CHECK-S27` against a PRD that *had* scoped concurrency out and correctly cost one sentence at `improvement`; `stage-exit-coverage` V-006 raised the same check against a silent PRD, was filed as a `gap`, and induced a full locking protocol that was later removed.
+   **A checklist item with no PRD position behind it is a PRD gap, not a design to invent.** Several checks are deliberately conditional — `CHECK-S27` ("Concurrent access scenarios are addressed **if relevant**") is the clearest, and the same shape appears for performance, observability, and security checks that defer to the PRD. When such a check fires and the PRD takes **no position** on the concern, the finding is that the *requirements* are silent. Report it as an `improvement` (or a `gap` against the PRD, in `prd` mode) whose suggested fix is to **record the position** — including "out of scope, single writer assumed", which is a complete answer. Do **not** specify a mechanism to satisfy the check: a verifier that answers an open requirements question by designing a protocol converts a one-sentence PRD amendment into a foundational change that no requirement asked for, at the stage where it is least visible. Precedent both ways: `epic-orchestration` V-008 raised `CHECK-S27` against a PRD that *had* scoped concurrency out and correctly cost one sentence at `improvement`; `stage-exit-coverage` V-006 raised the same check against a silent PRD, was filed as a `gap`, and induced a full locking protocol that was later removed. For **forge state writes** specifically, the standing owner decision is recorded in `references/decisions/single-writer-threat-model.md` (issue #180): single writer assumed, detection-not-locking — answer `CHECK-S27` by citing it (or the per-feature requirement restating it), never by designing a mechanism.
 
    Relatedly, **do not let `improvement` fall out of use.** A report containing only `gap`/`error`/`inconsistency` usually means observations that should have been `improvement` were promoted into must-fix findings. If a finding would not block implementation, it is an `improvement` — say so.
 3. Exact location (file + section)

package/adapters/claude/skills/forge-verify/references/decisions/single-writer-threat-model.md

@@ -0,0 +1,53 @@
+# Decision: single-writer state model — detection welcome, locking out of scope
+
+**Decided:** 2026-08-08 · **Issue:** [#180](https://github.com/garygentry/feature-forge/issues/180) · **Owner:** repository owner
+
+## The decision
+
+Concurrent multi-session mutation of forge state is **out of scope**. Every forge state
+writer — `forge-session.py` (`.pipeline-state.json`) and `epic-manifest.py`
+(`epic-manifest.json`, `.epic-state.json`) — assumes a **single writer**. Atomicity
+(sibling temp file → flush/fsync → `os.replace`) protects against an *interrupted* write;
+it is not, and is not intended to be, mutual exclusion between simultaneous writers.
+
+The standing posture is **detection, not locking**:
+
+- **No locking mechanism will be added on the authority of a verification finding or a
+  single feature's spec.** If concurrent multi-session use ever becomes a supported
+  workflow, that is a product decision needing its own PRD — and a state lock alone would
+  be false comfort, because git operations, the two-commit provenance protocol, and
+  adapter regeneration are equally unsynchronized.
+- **Cheap, opportunistic detection is welcome if someone proposes it** — e.g. an
+  epic-root writer re-reading and failing loudly on an unexpected `revision` before
+  `os.replace`. It is an explicitly optional hardening, not a requirement, and nothing is
+  scheduled.
+
+## Accepted residual risk
+
+Everything in-session is sequential, and member `.pipeline-state.json` files are
+disjoint. The one real exposure: two sessions working **different members of one epic**
+share two epic-root files, `epic-manifest.json` and `.epic-state.json`. A lost
+read-modify-write increment of the manifest's `revision` can leave it unchanged after a
+semantic mutation landed, and the freshness comparison then classifies a **stale epic
+verification as `fresh`** — a correctness consequence, not just a lost edit.
+
+This is **accepted and documented**, not mitigated. Anyone who hits it has this document
+to cite; do not run two concurrent sessions against members of the same epic.
+
+## What this means for verification (CHECK-S27)
+
+When `CHECK-S27` ("Concurrent access scenarios are addressed if relevant") fires against
+forge state writes, the answer is a **citation, not a design**: cite this document (or
+the per-feature requirement restating it, e.g. `stage-exit-coverage` `REQ-REL-04`,
+`epic-orchestration` `REQ-ROBUST-03`). A PRD silent on concurrency gets a one-sentence
+position recorded; a verifier must never answer the silence by specifying a mechanism.
+
+## Provenance
+
+Both precedents ran the same check; the only variable was whether the PRD had a position:
+
+- `stage-exit-coverage` V-006 raised CHECK-S27 against a silent PRD, was filed as a
+  `gap`, and induced a ~140-line portable lock protocol answering no requirement. It was
+  removed (`302c93f`), and the PRD now records the position as `REQ-REL-04`.
+- `epic-orchestration` V-008 raised the same check against a PRD that had scoped
+  concurrency out (`REQ-ROBUST-03`), was filed as an `improvement`, and cost one sentence.

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

@@ -2,6 +2,36 @@
 
 Loaded by the **parent orchestrator** role of `forge-verify` at Step 4 (write the findings document) and Step 6 (epic-mode state write). The `forge-verifier` leaf subagent MUST NOT load this file — it holds only orchestrator-facing material (see `SKILL.md` → "Which role are you?").
 
+## Truncated Verifier Returns (Synthesize gate)
+
+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
+plausible-sounding opening sentence ("I'll start by loading the pipeline state…") while
+the recovered digest contained two BLOCKER findings. The failure is quiet precisely
+because the returned line reads like progress, not like an error.
+
+**Gate every verifier return before using it.** A return is a **non-answer** when it
+lacks the report structure — no `# Verification Report:` header, no `## Findings`
+section, no `Checks Executed:` line (for a skeptic dispatch: no per-finding
+CONFIRMED/REFUTED verdicts). Length is the tell: a real report is rarely under ~20
+lines; a one-or-two-sentence return from a run that did real work is a dropped digest,
+not a clean result.
+
+On a non-answer:
+
+1. **Resume, don't re-run.** `SendMessage` to the returned agent id: *"Your run ended
+   without returning the report. Return your FINAL report now, from the work you
+   already did, in the Output Format — your final message is the only thing I
+   receive."* The agent replays from its transcript and typically returns the full
+   digest with zero further tool calls.
+2. **Re-dispatch** the same prompt only if the resume also fails or the id is gone.
+3. **Never** synthesize findings from a truncated return, write a findings document
+   around it, record a verify result (`state-verify`) from it, or present it to the
+   user as the verification outcome. A gate decision made on a dropped digest is the
+   #183 failure mode: BLOCKER findings silently converted into a pass.
+
 ## Findings Document Template (Step 4)
 
 Write findings to `{specsDir}/{feature}/.verification/VERIFY-{mode}-{YYYY-MM-DD}.md`

package/adapters/claude/skills/forge-verify/references/shared-conventions.md

@@ -36,7 +36,7 @@ I found that the codebase uses React and TanStack Router.
 
 ### Decision Support: Help the User Choose
 
-When an `AskUserQuestion` carries substantive options (a real choice — not a trivial yes/no confirmation), do not just list them. The interview stages have already done codebase research and integration analysis; surfacing that synthesis at the decision moment is the whole point. For every such question:
+When a question posed through `AskUserQuestion` carries substantive options (a real choice — not a trivial yes/no confirmation), do not just list them. The interview stages have already done codebase research and integration analysis; surfacing that synthesis at the decision moment is the whole point. For every such question:
 
 - **Lead with a recommended option.** Place it first and label it `(recommended)` (matching the `AskUserQuestion` "(Recommended)" convention).
 - **Put the trade-off in each option's `description`.** Say why you'd pick it and what you give up versus the alternatives — the cost, not just the benefit.
@@ -53,6 +53,17 @@ For genuinely comparable artifacts (competing module structures, two code snippe
 
 The **Branch Setup** block below is the reference pattern: a strong recommendation as the first option, rationale inline, the alternative still available, never a hard-stop.
 
+## Stage Review Gate
+
+Every authoring stage ends its "Review with User" step in exactly one of two shapes. Which shape a stage uses is declared **here, once** — each stage's review step points at this block by title, and the shape is never re-derived from the surrounding prose or inferred from how sibling stages behave (three stages block and one does not; majority-shape inference is precisely the failure this block exists to prevent):
+
+- **Blocking review (gate).** The stage presents the artifact and collects feedback through `AskUserQuestion`; it does **not** proceed until the user answers, iterating until they confirm. Stages: **forge-1-prd** (Step 5), **forge-2-tech** (Step 6), **forge-3-specs** (Step 6).
+- **Non-blocking review (invitation).** The stage states the artifact is ready and invites adjustments **as a statement, not a question** — and then **proceeds to the next step in the same turn unless the user asks for changes**. The invitation obliges the agent to *continue*: emitting the invitation sentence and stopping treats the non-gate as a gate and strands the stage `in-progress` with its completion step unrun — a defect, not caution. Stage: **forge-4-backlog** (Step 6).
+
+**Why the shapes differ.** A blocking review guards an artifact whose content was just authored from open-ended interview or synthesis — the user is the only authority on "complete", so the stage must wait. forge-4's backlog is *derived* from specs the user already approved, was planned interactively in its Step 3, and is machine-validated in its Step 5; a second hard gate would re-ask a settled question, and the loop never launches without forge-5-loop's own Step 2d confirmation anyway. The invitation is a courtesy checkpoint, not an approval gate (removed deliberately in #78's consistency sweep).
+
+A stage that changes shape changes it **in this block first**; the per-stage pointer stays a pointer.
+
 ## Configuration Reading
 
 Read `forge.config.json` from the project root. If it doesn't exist, use defaults.
@@ -130,7 +141,7 @@ mkdir -p "<specsDir>"
 [ -f "<specsDir>/AGENTS.md" ] || cp "$R/references/templates/specs-hygiene/AGENTS.md" "<specsDir>/AGENTS.md"
 ```
 
-If the host is Claude (the `AskUserQuestion` tool is available), also ensure the Claude-framed variant:
+If the host is Claude (the Claude-native question tool is available), also ensure the Claude-framed variant:
 
 ```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')"
@@ -191,7 +202,7 @@ Pipeline state is written by the `state-*` verbs of `scripts/forge-session.py`
 
 If a `state-*` verb exits 2, surface the plain `Error:` line from stderr verbatim, do **not** proceed to the next step of the surrounding protocol, and do **not** hand-author the JSON as a workaround. The stage remains resumable because the entry stamp is already on disk — re-run the verb once the cause is fixed.
 
-**The eight `state-*` verbs.** `state-enter` (Stage-Entry Guard), `state-artifact` (incremental artifact tracking), `state-complete` (Git Commit Protocol), `state-branch` (Branch Setup and Branch Reconciliation), `state-note` (the Immediate Downstream Note below, and the optional completion note at stage closure), `state-decision` (deferred decisions), `state-ecr` (epic change requests), and `state-verify` (one `forge-verify-*` verification transition — below). The `--epic` member requirement and the exit-2 failure protocol above apply to **every** one of them, `state-verify` included; no verify entry is ever hand-authored.
+**The nine `state-*` verbs.** `state-enter` (Stage-Entry Guard), `state-artifact` (incremental artifact tracking), `state-complete` (Git Commit Protocol), `state-skip` (the deliberate forge-6-docs documentation skip — scoped to that one stage; writes `status: "skipped"` + `skippedAt`, refuses to erase a record of docs that exist, and is the only sanctioned writer of a skipped docs stage), `state-branch` (Branch Setup and Branch Reconciliation), `state-note` (the Immediate Downstream Note below, and the optional completion note at stage closure), `state-decision` (deferred decisions), `state-ecr` (epic change requests), and `state-verify` (one `forge-verify-*` verification transition — below). The `--epic` member requirement and the exit-2 failure protocol above apply to **every** one of them, `state-verify` included; no verify entry is ever hand-authored.
 
 ### `state-verify` — verification results and provenance
 
@@ -395,7 +406,7 @@ Invoke this block **at the head of any post-entry step that writes a stage artif
 
 1. **Proceed** when `stages.{stage}.status` is `"in-progress"` (this session's Entry Stamp — you are finishing the run you started) or absent/`pending`. Run the write / exit normally.
 
-2. **Detect-and-refuse** when ALL of these hold: `stages.{stage}.status ∈ {"complete", "stale"}` **AND** the stage's artifacts (incl. `TRACEABILITY.md` for forge-3-specs) exist on disk **AND** a `commitHash` is recorded for the stage **AND** you did **not** author this stage earlier in the current session. This is a stale/replayed continuation of an already-finished, committed stage. Do **not** overwrite the artifact or re-run the exit. Route instead to the **Stage-Entry Guard**'s *Re-authoring* path: surface the same `AskUserQuestion` warning ("A completed {stage} artifact already exists for '{feature}' (v{n}{, marked stale}). Continuing will create a new version. Proceed?"). Only on explicit confirmation re-enter from the Entry Stamp (the version bumps at exit); otherwise **stop** and report that the stage is already complete — cite the recorded `commitHash` and offer `/feature-forge:forge {feature}` to see true state.
+2. **Detect-and-refuse** when ALL of these hold: `stages.{stage}.status ∈ {"complete", "stale"}` **AND** the stage's artifacts (incl. `TRACEABILITY.md` for forge-3-specs) exist on disk **AND** a `commitHash` is recorded for the stage **AND** you did **not** author this stage earlier in the current session. This is a stale/replayed continuation of an already-finished, committed stage. Do **not** overwrite the artifact or re-run the exit. Route instead to the **Stage-Entry Guard**'s *Re-authoring* path: surface the same warning via `AskUserQuestion` ("A completed {stage} artifact already exists for '{feature}' (v{n}{, marked stale}). Continuing will create a new version. Proceed?"). Only on explicit confirmation re-enter from the Entry Stamp (the version bumps at exit); otherwise **stop** and report that the stage is already complete — cite the recorded `commitHash` and offer `/feature-forge:forge {feature}` to see true state.
 
 When you cannot confirm you authored the current run, treat it as a replay and refuse: a false refuse costs one confirmation click; a false proceed overwrites a committed artifact and re-churns a stage version. `--force` follows Force Mode (skip the gate, treat as a deliberate re-author).
 

package/adapters/claude/skills/forge-verify/references/stage-exit-protocol.md

@@ -47,8 +47,8 @@ an epic member. Only the flags below are stage-specific; pass no others.
 |---|---|---|
 | `forge-0-epic` | `forge-0-epic` | `--next-feature "{member}"` when a concrete member exists |
 | `forge-1-prd` … `forge-4-backlog` | that stage's own id | none beyond identity/capability |
-| `forge-5-loop` | `forge-5-loop` | `--outcome` — one of `complete`, `partial`, `blocked`, `needs-human`, `deferred` |
-| `forge-6-docs` | `forge-6-docs` | `--outcome` — `complete` or `blocked` |
+| `forge-5-loop` | `forge-5-loop` | `--outcome` — one of `complete`, `partial`, `blocked`, `needs-human`, `deferred`, `resolved`; optional `--cause dependency-starvation` with `--outcome partial` |
+| `forge-6-docs` | `forge-6-docs` | `--outcome` — `complete`, `blocked`, or `skipped` (deliberate docs skip, persisted via `state-skip` before the exit; routes like `complete` with honest wording) |
 | direct `forge-verify` | `forge-verify` | `--owner direct`, `--outcome` (`passed`, `findings`, `skipped`, `failed`), and served-stage metadata |
 | nested `forge-verify` | `forge-verify` | `--owner nested`, plus the same outcome and served-stage metadata |
 | direct/nested `forge-fix` | `forge-fix` | the matching `--owner`, a `FixOutcome` (`no-findings`, `decisions`, `failed`, `applied`, `reverified`, `reverify-findings`, `deferred`), and served-stage metadata |
@@ -100,9 +100,9 @@ Obey the DIRECTIVES it prints, in the consumption order this protocol fixes: sur
 
 The stamp is shown with `--host claude`; the adapter build substitutes `pi`/`generic` per
 target, and §"Host and capability determination" below governs the value. The literal is
-deliberate — `scripts/build-adapters.py` keys its host translation on the exact string
-`--host claude`, and the stamp sites are compared byte-for-byte, so it is the one token in
-that line that is not a placeholder.
+deliberate — `scripts/build-adapters.py` keys its host translation on the exact canon
+value of that flag, and the stamp sites are compared byte-for-byte, so it is the one token
+in that line that is not a placeholder.
 
 ## Host and capability determination
 
@@ -110,8 +110,9 @@ Before the call, compute the two inputs independently. They are unrelated: **a h
 implies a capability**, and the script takes `--verify-capability` at face value.
 
 **`--host`** describes only the active adapter command surface — `claude`, `pi`, or
-`generic`. It selects command syntax (`/feature-forge:` vs `/skill:` vs host-neutral) and
-fresh-session wording (`/clear` vs `/new` vs neutral prose). Nothing else.
+`generic`. It selects command syntax (Claude's stage-command prefix vs Pi's `/skill:` vs
+host-neutral) and fresh-session wording (Claude's clear command vs Pi's `/new` vs neutral
+prose). Nothing else.
 
 **`--verify-capability interactive`** is passed only when **both** of these hold:
 
@@ -151,7 +152,8 @@ production successor** while verification is unresolved.
 - a capable Pi session is `--host pi --verify-capability interactive`, and receives the
   same logical gate a capable Claude session does;
 - Pi without a dispatchable verifier is `--host pi --verify-capability manual`;
-- a Claude session that cannot dispatch is `--host claude --verify-capability manual`.
+- a Claude session that cannot dispatch keeps the Claude host value with
+  `--verify-capability manual`.
 
 Interactive gate options keep their explicit labels, their recommended default, and their
 one-line trade-off descriptions (below). The manual path prints the verify command as the
@@ -199,6 +201,44 @@ second sentinel-terminated block **inside** an outer stage's exit, breaking the
 exactly-one-terminal-block rule — and the canon guard cannot catch it, because both
 wordings legitimately appear in the same file. Judge the token, not the phrasing.
 
+## Caller-side resumption: the declared resume point
+
+The `owner:` token and `terminalOwnedBy` arbitrate who prints — but they specify only
+the **callee** side: a nested skill stays quiet and returns its structured result. This
+section is the reciprocal, caller-side half of that contract.
+
+**On a sub-skill's return, the caller re-owns the terminal.** Every closing instruction
+in the callee's own body — its report-and-stop posture, its "confirm the result" close,
+its own next-steps habits — is void for this turn. The caller resumes at its **declared
+resume point**, in the same turn, and its remaining steps run to its own terminal.
+
+**Every Skill-tool delegation site declares, at the invocation, what happens on
+return.** Suppression without resumption is the failure mode this section closes: the
+callee's closing posture is the freshest instruction in context while the caller's next
+step is the oldest, so an undeclared return silently ends the run one layer too early —
+the caller's remaining steps (validation, state writes, commit, stage exit) dropped,
+with no error surfaced. An implicit "let it run to its natural stopping point" is that
+bug spelled politely, and it is banned on delegation sites. A site takes one of two
+declared postures:
+
+- **Delegate-and-resume** — the callee is a sub-step of the caller: the site names the
+  caller's own step that control returns to, and the caller continues there in the same
+  turn. The callee never owns the caller's terminal. The worked instance is
+  `forge-4-backlog` Step 4's **Return contract** for the `author-backlog` delegation
+  (control returns at Step 5; the sub-skill's direct-invocation posture — its approval
+  gate and its validate-and-confirm close — is explicitly disapplied on the delegated
+  path). New delegation sites follow that pattern rather than re-deriving it.
+- **Terminal handoff** — the caller's job ends at the invocation and the invoked skill
+  owns the terminal from there on (e.g. the navigator's `autoInvokeNextStage`
+  "continue in this session" advance into the next production stage). Declaring the
+  handoff is what keeps it distinct from an accidental drop.
+
+Scope: this contract governs **Skill-tool delegation within one session**. The
+truncated-verifier-return guard (forge-verify's `findings-template.md`, "Truncated
+Verifier Returns") is a different mechanism — it polices what an **Agent-dispatched
+subagent's** return payload must contain, not where a caller resumes. A dispatch site
+can be subject to both; satisfy each on its own terms.
+
 ## Directive consumption order
 
 `stage-exit` emits a DIRECTIVES object and (for a direct owner) a NEXT-STEPS block. The
@@ -299,7 +339,11 @@ first:
   a stage's artifact commit.
 - **Skip for now** — go straight to the NEXT-STEPS block without verifying. Record this
   stage's verify status as `skipped` in pipeline state (via `state-verify`, never by hand)
-  **only** on an explicit skip — a skip does not go stale.
+  **only** on an explicit skip — a skip does not go stale. Exception: if the existing
+  entry records `passed` or `findings-applied` (a resolved result whose freshness has
+  merely lapsed), write **nothing** — `state-verify` refuses to demote a resolved status
+  to `skipped` (#203), and the recorded result stands on its own; the user's decline is
+  honored by simply not re-verifying.
 
 **Advancement is allowed only after a pass, or after an explicit skip has been
 persisted.** Choosing to stop, or losing the interaction, produces no advancing terminal
@@ -358,8 +402,8 @@ directive is informational — you do **not** re-derive the wording:
   unchanged; the block appends a non-blocking reminder line ("You also flagged N epic
   change(s) to reconcile when convenient …"). This is *finish-then-edit*.
 
-Either way the added lines are host-neutral (no literal `/clear`) and sit **above** the
-sentinel; just print the NEXT-STEPS block verbatim as always.
+Either way the added lines are host-neutral (they name no fresh-session command) and sit
+**above** the sentinel; just print the NEXT-STEPS block verbatim as always.
 
 ### Deferred decisions — do not solicit next-stage decisions at this exit