| name | self-review-impl-architect |
| description | Authoritative tier + prompt-pack handoff for the Architect persona-switch self-critique round in the IMPLEMENTATION phase. PURE data; the Architect persona reasoning is in the implementation-agent's prompt. Returns the OKR's frozen governanceTier (passed in from the landing-issue Hatter Tag) + computed max_auto_rounds + should_proceed gate + the architect-review.md prompt pack contents read from .cheshire/prompts/implementation/ in the target repo. |
| version | 0.1.0 |
| purity | pure-data |
| runtime | research-runner |
| command | npx @maintainabilityai/research-runner@~0.1.64 skill-self-review-impl-architect |
Self-Review Impl Architect Skill
Hands the implementation-agent the authoritative inputs for the Architect persona's self-critique step. The agent MUST call this skill once per round (per the implementation-agent.agent.md contract) before writing the ### Self-review — Architect (round <N>) block. No LLM inside — the agent does the actual reasoning.
Why this skill exists (Codex-r4 Bug 2)
The implementation-agent template originally told the agent to call self-review-code-architect (the WHAT phase skill). That skill reads the mesh's okrs/<id>/okr.yaml and looks up the actions[] entry by runId to source the authoritative governanceTier. The contract breaks for the implementation-agent because:
- The impl agent runs INSIDE the TARGET REPO (no
MESH_PATH).
- The impl agent's
runId is IMPL-* — not present in okr.yaml's actions[] (those are WHY/HOW/WHAT run ids).
Calling self-review-code-architect from the target-repo context returns either okr-not-found (no mesh) or action-not-found: no actions[] entry with runId=IMPL-.... The audit chain would lack the self-review skill_call event entirely.
This skill mirrors the same data-only contract but adapts to the impl context:
- prompt pack is read from
<repo>/.cheshire/prompts/implementation/architect-review.md (the Cheshire scaffold installs a starter pack on first fan-out into a target repo);
tier is passed inline by the agent (the landing issue's Hatter Tag carries the OKR's frozen tier);- the
runId MUST start with IMPL- — defense-in-depth against accidental WHAT-phase routing (mirrors the audit-emit-event phase/runId guard, Codex-r4 Bug 3).
Inputs (stdin JSON)
| Field | Type | Required | Description |
|---|
okrId | string | yes | The OKR id, e.g. OKR-2026Q2-IMDB-001-celeb-api. Used for labeling — NOT to read mesh state. |
runId | string | yes | MUST start with IMPL-. Your implementation_run_id. Anything else returns runid-not-impl. |
round | integer | yes | The round number the agent is entering (1, 2, 3 — bounded by max_auto_rounds). |
tier | string | yes | Authoritative governance tier from the landing issue's Hatter Tag continuation (autonomous / supervised / restricted). The OKR's frozen tier — do not infer this from any other source. |
Outputs (stdout JSON)
| Field | Type | Description |
|---|
ok | boolean | true on success. false with reason when runId does not start with IMPL-. |
persona | "impl-architect" | Constant — confirms which skill ran. |
phase | "implementation" | Constant — pins the phase for downstream UI/metric extractors. |
tier | "autonomous" | "supervised" | "restricted" | "unknown" | Echo of the input, lowercased. |
maxAutoRounds | integer | Derived from tier: autonomous=3, supervised=2, restricted=0 (unknown→0). |
round | integer | Echoes the input. |
shouldProceed | boolean | true iff tier != restricted AND round <= maxAutoRounds. Use this to decide whether to write a ### Self-review — Architect (round N) block or a ### Self-review — Skipped block. |
promptPack | string | Full text of <repo>/.cheshire/prompts/implementation/architect-review.md. Empty when the file isn't present (promptPackFound: false). |
promptPackPath | string | Repo-relative path the skill read from. |
promptPackFound | boolean | true if the prompt-pack file exists in the target repo. |
payload emitted into the audit chain (via auditMetadata): {persona, phase, tier, max_auto_rounds, round, should_proceed, prompt_pack_path, prompt_pack_found}. The full promptPack body is NOT in the chain (would bloat events) — only the path + presence flag.
Invocation
echo '{"okrId":"OKR-X","runId":"IMPL-2026-05-28-celeb-api-abc123","round":1,"tier":"supervised"}' \
| npx -y @maintainabilityai/research-runner@~0.1.64 skill-self-review-impl-architect
Error contract
{ok: false, reason: 'bad-input: ...'} on schema failures (missing required fields). {ok: false, reason: 'runid-not-impl: ...'} when the run id does not start with IMPL- (the impl phase contract — pass parent_run_id to self-review-code-architect if you actually want the WHAT-phase skill).
