lidge-jun/codexclaw / plugins/codexclaw/skills/pabcd
cxc-pabcd
VerifiedUse for Plan-Audit-Build-Check-Done work. Triggers: PABCD, plan this, 기획, 단계별로, 요구사항 정리.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 5.2 KB |
| agents/openai.yaml | config | 147 B |
| references/catalog-discovery.yaml | config | 15.5 KB |
| references/delegation.md | markdown | 19.4 KB |
| references/dispatch-surfaces.md | markdown | 13.9 KB |
| references/implementation-units.md | markdown | 3.7 KB |
| references/loop-engineering.md | markdown | 9.3 KB |
| references/optimization.md | markdown | 3.9 KB |
| references/phase-audit.md | markdown | 8.4 KB |
| references/phase-check.md | markdown | 4.7 KB |
| references/phase-control.md | markdown | 9.7 KB |
| references/phase-plan.md | markdown | 6.5 KB |
| references/plan-output.md | markdown | 3.9 KB |
references/phase-audit.md
View on GitHub# Audit phase
2. **A — Audit**: Adversarial, read-only review of the plan against the real codebase. Dispatch an independent read-only reviewer using the exposed spawn schema (`agent_type:"reviewer"` when available; otherwise supported `agent_type:"explorer"` with `CXC-ROLE: reviewer` before `TASK:`; when the schema has no `agent_type`, use the same header and explicit read-only scope in supported `message` or text-`items`, per the [live-schema transport owner](delegation.md#live-tool-schema-and-role-transport)). Logical scope does not claim a native sandbox. Dispatch this reviewer — even a small/mini-model one — to challenge assumptions, find blockers (rollback gaps, missing callers, phantom constants), and verify references. A small/mini-model reviewer earns its place on fact checks — missing callers, wrong paths, phantom constants, stale references; give its whole-design rewrite suggestions little weight, and for substantial design work prefer the strongest reviewer available (AUDIT-REVIEWER-ROLE-01). For each conditional path the plan adds, the reviewer also asks: is the trigger reachable at all from states the system actually visits (callers exist, preconditions can co-occur, upstream code does not consume the trigger first), and does the plan name its activation scenario (C-ACTIVATION-GROUNDING-01)? An unreachable-by-construction branch is a plan blocker, not a C-phase discovery. The reviewer also checks: new devlog phase documents use the numbered lexicographic filename convention; bare-named or research/implementation-mixed docs are a FAIL (LEXICO-SPLIT-01). Multi-phase units satisfy DIFFLEVEL-ROADMAP-01: every roadmap phase has a diff-level decade doc (no outline-only or missing phases), and the phase map is dependency-ordered, not effort-bucketed (PHASE-SPLIT-01). **Audit loop (STRICT, AUDIT-LOOP-01):** A is a loop — audit -> synthesize -> amend plan -> re-audit — not a single round. Exit A>B only when the MAIN agent judges the round **pass** (reviewer approved) or **near-pass**: every High/Critical blocker was folded into the plan as a concrete amendment or explicitly rebutted with recorded rationale; a High/Critical rebuttal names a checkable basis when one exists — a command actually run with its output tail, or the exact path:line that shows the premise is false — and a prose-only "it is not so" does not qualify for near-pass (REBUTTAL-BASIS-01, agent-followed; the gate does not parse it), and only non-blocking residuals remain (`GO-WITH-FIXES; 2 blockers folded back` qualifies — the main agent is the judge, not a string parser). A FAIL round never exits: apply REVIEW-SYNTHESIS-01 (loop-engineering.md §11.3), amend the plan, and re-audit with the SAME reviewer (V2 `followup_task` to its task_name or V1 `send_input` to its agent_id; DISPATCH-ACTOR-01); a re-audit round looks at whether each prior blocker is closed and what the amendment itself broke; a new Low/Medium finding outside that becomes a non-blocking residual in `auditResidual` instead of a FAIL, while a new High/Critical finding is still folded into the plan or rebutted under REBUTTAL-BASIS-01 before A>B, or the work returns to P (AUDIT-REAUDIT-SCOPE-01 — this is what lets the loop converge instead of trading one round of nits for the next); LOOP-REPAIR-01 bounds the loop — after 3 failed rounds return to P with a changed plan (HITL may return to Interview). The dispatch packet explicitly names `$codexclaw:cxc-dev-code-reviewer` AND `$codexclaw:cxc-search` (reference/version/external-claim verification rides the search ladder) and instructs the reviewer to end with a normalized final line `VERDICT: PASS | GO-WITH-FIXES (blockers=N) | FAIL` plus numbered blockers. No code changes. The `A>B` attest structurally requires `auditOutput` (the pasted tail of the reviewer's verdict) plus `auditVerdict` (`pass|near-pass|fail` — the MAIN agent's own judgment of the round); `near-pass` additionally requires `auditResidual` naming each residual blocker and its disposition (folded/rebutted). A declared `fail` never advances, and a pasted tail whose final verdict line says FAIL is rejected regardless of the claimed judgment. Still a form-only bar: the gate cannot verify the paste's provenance, so faithful execution (really dispatching the reviewer, really looping) remains the agent's obligation.
**Plan-rule checks (PLAN-VERIFIER-REAL-01 / PLAN-FIELD-CHAIN-01 / PLAN-BYPASS-NAMED-01).** The reviewer additionally verifies, and any one of these failing is a blocker: (a) every verifier command the plan names actually exists AND reads the change target — the reviewer RUNS it rather than trusting the plan; (b) each new field/enum value has its full creation -> serialization -> deserialization -> consumer chain enumerated, with `N/A + reason` where a stage does not apply; (c) when several documents reference a shared type, the field NAMES match, not just the concept; (d) each document's header dependency declaration matches the types its body actually uses; (e) any plan adding enforcement records the five bypass fields (tier / executing surface / known bypass / residual risk / wording downgrade) and either names the final enforcement layer or states `none`.
**Verification is not pinned to A (LEAN-REVIEW-01, 260818).** `cxc review-round open` still exists, and a verdict a reviewer's exit records is still honoured: it cannot be contradicted by your attest, spent after a re-plan, or spent on a plan that changed since. What changed is that an open round no longer BLOCKS `A>B`. It used to, and that produced a deadlock — every reason the SubagentStop observer did not fire (a matcher that missed the runtime's role vocabulary, a reviewer whose closing lines did not parse, a reinstall that moved `PLUGIN_ROOT` out from under a live session) became a cycle that could never leave A, whose only escape was hand-feeding the hook its own payload. A gate whose normal recovery is forging its own input is not a gate.
Dispatch reviewers wherever they actually help — plan audit at A, implementation lanes at B, verification lanes at C — instead of treating A as the one phase that owns review. Open a round when you want the verdict bound to the plan hash and this cycle; otherwise attest and move. Both are honest; only claiming a reviewer ran when none did is not.
When the verdict is FAIL, fold-back follows REVIEW-SYNTHESIS-01 (loop-engineering.md §11.3): synthesize root causes and accept/rebut decisions before re-planning or re-dispatching the reviewer.
- **Reviewer reuse across repair rounds (pointer):** blocker-closure re-verification
rounds reuse the SAME reviewer — V2 `followup_task` to its task_name (triggers a
turn when idle; `send_message` is context-only) or V1 `send_input` to its agent_id —
passing the synthesis plus a change-diff summary so the reviewer keeps its context.
The final C adversarial gate (or any contaminated reviewer) gets a fresh reviewer or
a direct independent audit instead. Reuse is the default; the round count alone does
not make a reviewer contaminated. Normative lifecycle rules: DISPATCH-ACTOR-01 /
DISPATCH-RETIRE-01 in `structure/20_pabcd_dispatch_doctrine.md` §3
(repository-only provenance, not an installed prerequisite).
## Architect recheck after audit amendments
Keep the independent reviewer and its audit loop. Reinvoke the plan's architect only
when an amendment changes a documented design decision: module responsibility,
data structure, interface or execution flow. Main names the decision ID, before/after
and reason, updates the executable plan, and requests reflection from the SAME architect
before completing A. Text edits or test clarification alone do not trigger a call.
If test work changes an interface or flow decision, that decision change does.
Record the reflection and main disposition separately, then re-audit the amended plan
with the reviewer. Lifecycle and failure handling: [delegation](delegation.md#architect-context-and-routing).
Prefer receiving the reviewer result before requesting architect revisions; if opening a new
review round, do so after reflection. An unsigned architect exit overlapping an in-flight A
round can produce `review_signoff_unparsed` in the existing review observer. That is a
diagnostic about missing reviewer sign-off, not architect failure, A approval or a gate
verdict. Record its cause if it occurs. Do not manufacture `LAUNCH`/`VERDICT` reviewer
sign-off for the architect; its `ALIGNED`/`MISALIGNED` result has a different purpose.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| rea-analysis-optimizationmorluto/reaOptimize REA providers and shared analysis workflows for agent context, latency, CPU, and memory while preserving correctness and Evidence. Use for analysis bottlenecks and cache or query design; CI/test-suite optimization belongs to rea-ci-tests-audit. | Stars: 471k | Updated | Claude CodeCodex | Verified | Install |
| rea-ci-tests-auditmorluto/reaAudit REA CI and test suites for correctness, fixture drift, false passes, and measured performance; use for suite or workflow investigations and efficiency changes, rather than ordinary feature testing or a single test failure. | Stars: 471k | Updated | Claude CodeCodex | Verified | Install |
| rea-tool-designmorluto/reaAudit, investigate, design, or change REA investigation tools, CLI/MCP contracts, provider capabilities, and Evidence semantics. | Stars: 471k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-auditDietrichGebert/ponytailQuality audit of the whole repo: bugs, security, real load, missing tests, speed, and what to delete. Most important first. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |