agentsclimarketplace
Sign in

Review

Skill GrinRus/ai_driven_dev/skills/review

Runs review-stage validation for scope changes, findings, and follow-up task derivation. Use when review stage needs verdict and handoff tasks. Do not use when the request is direct implementation execution in `implement` or QA validation in `qa`.From its SKILL.md

Install
npx -y skills add GrinRus/ai_driven_dev --skill review

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 5 stars5 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

7.2 KB, as published. Nobody here has run it

Follow feature-dev-aidd:aidd-core and feature-dev-aidd:aidd-loop.

Steps

  1. Inputs: resolve active <ticket>/<scope_key> and verify loop/readmap artifacts required for review stage.
  2. Stage-chain policy: execute via canonical stage-chain orchestration through stage runtime entrypoint; internal preflight/postflight are orchestration details and not operator commands.
  3. Manual write/create of stage.review.result.json is forbidden; stage-result files are produced only by stage-chain postflight. [AIDD_LOOP_POLICY:MANUAL_STAGE_RESULT_FORBIDDEN]
  4. Runtime-path safety: use only runtime commands from this skill contracts. If stdout/stderr contains can't open file .../skills/.../runtime/..., stop with BLOCKED runtime_path_missing_or_drift; do not invent alternate filenames and do not retry guessed commands.
  5. Test command safety: do not run raw build/test commands directly in review stage orchestration. Manage test requirement via python3 ${CLAUDE_PLUGIN_ROOT}/skills/review/runtime/reviewer_tests.py and existing aidd/reports/tests/** evidence. If you detect cwd mismatch (command path missing or command cannot run from selected cwd), record blocker tests_cwd_mismatch and stop repeated retries.
  6. Read order after stage-chain preflight artifacts: readmap.md -> loop pack -> review pack (if exists) -> rolling context pack; do not perform broad repo scan before these artifacts.
  7. Run subagent feature-dev-aidd:reviewer.
  8. Orchestration: produce review artifacts with python3 ${CLAUDE_PLUGIN_ROOT}/skills/review/runtime/review_report.py, python3 ${CLAUDE_PLUGIN_ROOT}/skills/review/runtime/review_pack.py, python3 ${CLAUDE_PLUGIN_ROOT}/skills/review/runtime/reviewer_tests.py, and python3 ${CLAUDE_PLUGIN_ROOT}/skills/aidd-flow-state/runtime/tasks_derive.py; then Fill actions.json for aidd/reports/actions/<ticket>/<scope_key>/review.actions.json strictly as aidd.actions.v1 (schema_version, allowed_action_types, canonical type + params) with action types only from {tasklist_ops.set_iteration_done, tasklist_ops.append_progress_log, tasklist_ops.next3_recompute, context_pack_ops.context_pack_update}, and validate via python3 ${CLAUDE_PLUGIN_ROOT}/skills/review/runtime/review_run.py.
  9. Canonical stage-chain: internal preflight -> stage runtime -> actions_apply.py/postflight -> python3 ${CLAUDE_PLUGIN_ROOT}/skills/aidd-flow-state/runtime/stage_result.py; it must produce aidd/reports/loops/<ticket>/<scope_key>/stage.review.result.json. [AIDD_LOOP_POLICY:CANONICAL_STAGE_RESULT_PATH]
  10. Non-canonical stage-result path under skills/aidd-loop/runtime/ is forbidden (treat as prompt-flow drift). [AIDD_LOOP_POLICY:NON_CANONICAL_STAGE_RESULT_FORBIDDEN]
  11. Output: return one terminal review-stage payload after evidence-first evaluation, with findings summary and next action or handoff.

Command contracts

python3 ${CLAUDE_PLUGIN_ROOT}/skills/review/runtime/review_run.py

  • When to run: as canonical review stage runtime before postflight.
  • Inputs: ticket, scope/work-item context, and reviewer findings/actions payload.
  • Outputs: validated review artifacts and stage status payload.
  • Failure mode: non-zero exit on invalid review contracts or missing prerequisites.
  • Next action: fix findings/actions inputs and rerun runtime validation.

python3 ${CLAUDE_PLUGIN_ROOT}/skills/review/runtime/reviewer_tests.py

  • When to run: when reviewer must set/clear test requirement marker for the current scope.
  • Inputs: --ticket, optional --scope-key, --work-item-key, and --status (required|optional|skipped|not-required).
  • Outputs: reviewer marker at aidd/reports/reviewer/<ticket>/<scope_key>.tests.json.
  • Failure mode: invalid status or unresolved ticket/scope context.
  • Next action: fix marker arguments/context and rerun once; do not replace with ad-hoc shell test loops.

python3 ${CLAUDE_PLUGIN_ROOT}/skills/aidd-docio/runtime/actions_apply.py

  • When to run: mandatory final step in stage-chain postflight after review actions are validated.
  • Inputs: --actions <path> and optional --apply-log <path>.
  • Outputs: applied actions plus progress/stage-result/status-summary artifacts.
  • Failure mode: apply failure, boundary guard failure, or summary generation failure.
  • Next action: inspect logs, fix blocking artifact/contract, rerun the stage-chain, and verify canonical stage result exists (aidd/reports/loops/<ticket>/<scope_key>/stage.review.result.json).

python3 ${CLAUDE_PLUGIN_ROOT}/skills/aidd-flow-state/runtime/stage_result.py

  • When to run: stage-chain postflight stage-result emission only (not operator/manual recovery command).
  • Inputs: canonical postflight payload (ticket, stage, result, scope-key, work-item-key, evidence links).
  • Outputs: canonical aidd.stage_result.v1 at aidd/reports/loops/<ticket>/<scope_key>/stage.review.result.json.
  • Failure mode: non-zero exit on missing required args or invalid stage-result contract fields.
  • Next action: fix postflight payload generation and rerun the stage-chain; do not switch to non-canonical loop runtime paths.

Notes

  • Review stage sets/verifies test requirement via reviewer marker and existing reports; avoid ad-hoc repeated shell retries.
  • Use the existing rolling context pack; do not regenerate it in loop mode (DocOps updates only).

Additional resources

  • Contract schema: CONTRACT.yaml (when: review artifacts or actions contract is ambiguous; why: verify required structure before rerunning runtime/postflight).

Keep looking

Skills are one crate of 627,801. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.