---

name: return-path-gate description: Audits SKILL.md files in plugins/ for return-path gate compliance. Detects fire-and-forget chains where a caller dispatches a downstream skill but does not gate its next step on the callee's verdict. Reports each chain as CLOSED or OPEN and outputs specific fix suggestions. Triggered by "chain audit", "return-path check", "fire-and-forget detection", "chain closure audit", "skill chain compliance". user-invocable: true allowed-tools: ["Read", "Bash", "Glob", "Grep"] model: sonnet

return-path-gate — Skill Chain Closure Audit

A skill chain is CLOSED when the callee's verdict folds back into the caller's decision logic. A chain is OPEN (fire-and-forget) when the caller dispatches a downstream skill but proceeds independently of the result.

This skill audits §Chains and §Done When sections across SKILL.md files to detect OPEN chains and prescribe specific closures.

Role Separation from harness-doctor

Dimension	harness-doctor	return-path-gate
What it checks	Structural completeness (files, references, line counts, drift)	Chain closure compliance (caller/callee verdict handoff)
Core question	"Are the required harness files present and consistent?"	"Does the caller wait for the callee's verdict?"
Failure mode detected	Missing CLAUDE.md, broken refs, complexity drift	Fire-and-forget chains, unverified CONDITIONAL_PASS conditions
Output	L1–L5 structural health report	CLOSED/OPEN chain table with fix prescriptions

---

Trigger Phrases

Phrase	Situation
"chain audit", "chain closure audit"	Full scan of all skill chains in plugins/
"return-path check", "return-path gate"	Explicit compliance audit
"fire-and-forget detection", "fire-and-forget chain"	Targeting known anti-pattern
"skill chain compliance"	Pre-flight check before pipeline execution
"does this skill wait for verdict"	Single-skill targeted check
"check chains in [skill-name]"	Scoped audit on a named skill
/return-path-gate	Explicit call

---

Audit Scope Options

Flag	Scope	When to use
--skill [name]	Single named skill	Targeted check during development
--all	All SKILL.md files under plugins/	Full compliance sweep
--pr	Skills changed relative to origin/main	Pre-PR gate mode (worktree-aware)
(default)	Modified + newly added SKILL.md files in working tree	Pre-commit gate mode

---

Execution Steps

Step 0. Determine Scope

# Default: modified and newly added SKILL.md files in working tree
{ git diff --name-only HEAD; git ls-files --others --exclude-standard; } | grep "SKILL\.md"

# --pr mode: PR-relative, worktree-aware (use for pre-PR audits)
git diff --name-only origin/main...HEAD | grep "SKILL\.md"

# --all: full scan
find plugins/ -name "SKILL.md" | sort

# --skill [name]: targeted
find plugins/ -path "*[name]/SKILL.md"

If 0 SKILL.md files are in scope, distinguish the two causes before stopping — a failed scope command and a genuinely empty scope both render as zero lines, and only one of them is skippable:

git rev-parse --git-dir >/dev/null 2>&1 \
  || { echo "HARNESS_ERROR: git unavailable — scope could NOT be measured (not an empty scope)"; exit 10; }

# Repo-liveness alone is NOT enough: a broken ref (e.g. missing origin/main in --pr mode) fails the
# scope command with rc=1 and empty output while rev-parse stays green. Capture the scope command's
# OWN exit status — before the grep, which legitimately returns 1 on no-match:
scope_out=$(git diff --name-only origin/main...HEAD 2>&1); scope_rc=$?
[ "$scope_rc" -ne 0 ] \
  && { echo "HARNESS_ERROR: scope command failed (rc=$scope_rc): $scope_out"; exit 10; }
printf '%s\n' "$scope_out" | grep "SKILL\.md"
# (same rc-capture shape applies to the default and --all scope commands)

With both controls green and genuinely 0 files in scope, output: "No SKILL.md files in scope — audit skipped (control: git rev-parse OK)." and stop. A HARNESS_ERROR is a distinct non-pass — it must never be reported as "audit skipped", because this skill is pipeline-conductor's Step 0.5 pre-flight and a tooling-down skip would render as a pass there.

---

Step 1. Extract Chain Declarations

For each SKILL.md in scope, extract:

1. §Chains section — every → [skill-name] reference (downstream dispatch)
2. §Done When section — completion conditions listed by the caller

Two-pass extraction (mandatory — do not conflate the two passes):

Pass A — Caller-wait signals (look for in §Done When of the caller skill):

"wait for"          ← waits for callee verdict before proceeding
"verdict received"
"folds back"
"gates on"
"callee returns"

These signals indicate the caller waits for the callee's verdict. They are the primary CLOSED gate.

Pass B — Callee-output signals (look for in §Done When of the callee skill):

Verdict: PASS | CONDITIONAL_PASS | FAIL | ESCALATE

These signals indicate the callee emits a structured verdict. They confirm the callee side is closeable. They do not indicate the caller waits — treat as callee-role evidence only and never conflate with Pass A.

Step 1 output:

## Step 1 — Chain Extraction

Skill: [skill-name]
  §Chains dispatches: → [callee-1], → [callee-2]
  §Done When (Pass A — caller-wait signals): [extracted text or "none"]
  [callee-1] §Done When (Pass B — callee-output signals): [extracted text or "none"]

---

Step 2. Classify Each Chain as CLOSED or OPEN

Apply the following decision rules per (caller → callee) pair.

CLOSED criteria:

1. (Mandatory) Caller §Done When has Pass A signals (caller-wait: "wait for", "verdict received", "folds back", "gates on", "callee returns")
2. (Mandatory) Callee §Done When has Pass B signals (callee-output: Verdict: PASS | CONDITIONAL_PASS | FAIL | ESCALATE)
3. (Advisory — not independently required) Caller §Chains uses the closed form: → Mandatory next ([callee]) when [condition]; verdict folds back into [step]

CLOSED when: Criteria 1 AND 2 both pass. Criterion 3 adds confidence but is not mandatory — a skill with informal §Chains language that still satisfies criteria 1+2 is CLOSED.

OPEN when: Criterion 1 OR criterion 2 fails:

• Caller §Done When has no Pass A signals, OR
• Callee §Done When has no Pass B signals

Highest-risk OPEN pattern — CONDITIONAL_PASS gap: Callee §Done When lists conditions under CONDITIONAL_PASS but caller §Done When has no mechanism to enforce resolution. Classify as OPEN + flag CONDITIONAL_PASS gap in basis.

Severity assignment (two dimensions — take the higher tier as final severity):

Caller tier:

Caller identity	Tier
Core pipeline skill (harvest-loop, steel-quench, apex-review, agent-composer, sim-conductor, pipeline-conductor)	HIGH
Diagnostic or gate skill (harness-doctor, phantom-quench, verify-bidirectional, return-path-gate)	MEDIUM
Utility or advisory skill (context-doctor, plugin-recommender, frontier-digest, etc.)	LOW

Callee consequence tier:

Callee output capability	Tier
Callee can output FAIL or ESCALATE verdicts	HIGH
Callee can output CONDITIONAL_PASS verdicts	MEDIUM
Callee is advisory only (no structured verdict)	LOW

Final severity = max(caller tier, callee consequence tier).

Step 2 output:

## Step 2 — Chain Classification

| Caller | → Callee | Status | Severity | Basis |
|---|---|:---:|:---:|---|
| [skill] | [callee] | CLOSED | — | Criteria 1+2 pass; criterion 3 advisory met |
| [skill] | [callee] | OPEN | HIGH | Caller §Done When: no Pass A signals |

---

Step 3. Generate Fix Prescriptions for OPEN Chains

For each OPEN chain, output a specific prescription. Use the Implementation Checklist from knowledge/shared/harness-core/return_path_gate.md as the fix template.

Core FH skill scope check: If the callee is a core FH skill (harvest-loop, steel-quench, sim-conductor, apex-review, agent-composer, pipeline-conductor), output a proposal-only note flagged for human review. Do not generate a standard callee-side prescription — changes to core FH skills have downstream consequences requiring deliberate review.

Tool limitation: allowed-tools does not include Write or Edit. All prescriptions are text output only — manual application required.

Fix prescription format:

### OPEN: [caller] → [callee] (Severity: HIGH/MEDIUM/LOW)

Required changes:

1. [caller]/SKILL.md §Chains:
   Replace: → [callee]
   With:    → Mandatory next ([callee]) when [condition]; verdict folds back into [step N]

2. [caller]/SKILL.md §Done When:
   Add: "Wait for [callee] verdict before proceeding to [next action]"

3. [callee]/SKILL.md §Done When:
   [If callee is a core FH skill] → PROPOSAL ONLY — flag for human review before modifying
   [Otherwise] Add structured verdict block:
   Verdict: PASS | CONDITIONAL_PASS | FAIL | ESCALATE
   Conditions: [list — only present when CONDITIONAL_PASS]
   Basis: [one-line rationale]

4. [callee]/SKILL.md §Chains (if not present):
   Add: "Returns verdict to [caller]"

Simplification guard: If callee already outputs a structured verdict (Pass B satisfied) but caller only lacks Pass A signals, output only caller-side changes (items 1 and 2).

---

Step 4. Summary Report

## return-path-gate Audit Complete

Scope: [--skill name | --all | --pr | changed files (default)]
Skills audited: N
Chains evaluated: N

Results:
  CLOSED: N chains
  OPEN:   N chains (HIGH: N / MEDIUM: N / LOW: N)

HIGH severity OPEN chains (requires immediate fix):
  - [caller] → [callee]

Next actions:
  - Apply prescriptions above for each OPEN chain
  - Re-run /return-path-gate --skill [name] after each fix to confirm CLOSED
  - HIGH severity chains block pipeline-conductor Step 0.5 pre-flight

---

Done When

Step 0 scope determined; a failed scope command reported as
  HARNESS_ERROR, never as empty scope                           — mandatory-pass
+ Step 1 two-pass extraction complete (Pass A caller-wait +
  Pass B callee-output evaluated separately)                    — mandatory-pass
+ Step 2 every (caller → callee) pair classified CLOSED or
  OPEN with two-dimension severity                              — judged (adversarial pairing:
  before ANY chain is reported OPEN, the ground-truth CLOSED
  pair from §Operating Notes must classify CLOSED in the same
  run — an instrument that cannot separate the known pair
  reports nothing)
+ Step 3 fix prescriptions output for each OPEN chain
  (core FH skills: proposal-only)                               — mandatory-pass
+ Step 4 summary report with CLOSED/OPEN counts output          — measured (counts)

Fix prescriptions are text output only (Write not in allowed-tools). Prescription application is manual and out of scope for this skill. Verification of applied fixes requires re-running /return-path-gate --skill [name].

Verdict: PASS (0 HIGH severity OPEN chains) | CONDITIONAL_PASS (MEDIUM/LOW severity OPEN chains only) | FAIL (1+ HIGH severity OPEN chains found)

---

Connected Skills

Situation	Connected Skill
Check harness structural completeness alongside chain closure	/harness-doctor
Verify phantom references in §Chains targets (callee skill actually exists)	/phantom-quench
Run chain audit as pre-flight before parallel dispatch	pipeline-conductor Step 0.5 calls this skill
Prescribe verdict format for callee skills missing structured output	/meta-prompt-builder
Chain OPEN finding becomes improvement candidate	/field-harvest

---

Operating Notes

• Read callee SKILL.md directly: Do not infer whether a callee outputs structured verdicts from memory. Always Read the callee's §Done When section.
• Two-pass extraction is mandatory: Pass A (caller-wait signals) and Pass B (callee-output signals) serve different purposes. Never conflate them — a callee-output signal in §Done When is not a caller-wait signal.
• Criterion 3 is advisory: §Chains formal language (→ Mandatory next ...) adds confidence but criteria 1+2 are the mandatory gate. A skill with informal §Chains language that still satisfies criteria 1+2 is CLOSED.
• CONDITIONAL_PASS gate is the highest-risk gap: A chain where CONDITIONAL_PASS conditions are listed by the callee but the caller has no enforcement path is OPEN even if other verdict paths fold correctly.
• Scope default is narrow by design: Default captures modified + newly added files (not just staged changes). Use --pr for PR-relative mode in worktrees, --all for periodic sweeps.
• Core FH skill prescriptions are proposals only: Edits to harvest-loop, steel-quench, sim-conductor, and other core skills require deliberate review — output proposal notes, not standard prescriptions.
• Reference pattern: knowledge/shared/harness-core/return_path_gate.md defines the canonical closed-loop structure and verified instances (apex-review → sim-conductor, agent-composer ↔ deliberation). These are the ground-truth CLOSED examples for calibrating classification — and that calibration is a Done When condition (Step 2's adversarial pairing), not an optional note.
• The Pass A signal list is a closed English keyword list — know its degrade direction: synonyms ("blocks until", "halts on") and non-English documentation will read as false OPEN. That direction is by design (over-report, never silent-close), but treat an OPEN verdict on unconventional wording as a candidate for human reading before prescribing a fix.