grinrus/aidd-loop
skills/aidd-loop/SKILL.md
Enforces loop-mode discipline for implement/review/qa (packs, scope, no questions). Use when loop stages require bounded scope and retry policy. Do not use when direct stage execution belongs to `implement`, `review`, or `qa` stage skills.
$ install --global
skillsauthnpx skillsauth add grinrus/ai_driven_dev aidd-loopInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
Security Scan Results
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
3
Scanners Passed
9
Scanners in report
Clean
TrivyContainer and dependency vulnerability scanner
95%
Clean
SemgrepStatic code analysis for vulnerabilities
95%
Clean
mcp-scan (Snyk)Model Context Protocol security validation
95%
Skipped
Snyk (dep)Open source security scanning
50%
Skipped
Socket.devSupply chain security analysis
50%
Skipped
VirusTotalMulti-engine malware detection
50%
Skipped
CrowdStrikeAdvanced threat intelligence
50%
Skipped
OSV-ScannerOpen Source Vulnerability database check
50%
Skipped
OWASP Dep-Check
50%
Last scanned: Apr 22, 2026, 8:05 PM16.1s23 files scanned
SKILL.md
- name:
- aidd-loop
- description:
- Enforces loop-mode discipline for implement/review/qa (packs, scope, no questions). Use when loop stages require bounded scope and retry policy. Do not use when direct stage execution belongs to `implement`, `review`, or `qa` stage skills.
- lang:
- en
- model:
- inherit
- user-invocable:
- false
Follow feature-dev-aidd:aidd-core for output contract and DocOps.
Loop discipline
- Read order: loop pack -> review pack (if any) -> fix_plan.json when verdict=REVISE -> rolling context pack.
- Excerpt-first: do not read full PRD/plan/tasklist/spec when the loop pack excerpt covers Goal/DoD/Boundaries/Expected paths/Size budget/Tests/Acceptance.
- No questions in loop-mode; record blockers/handoff instead.
- REVISE: reuse the same scope, follow fix_plan.json, and do not widen boundaries.
- Scope guard: out-of-scope -> WARN + handoff; forbidden -> BLOCKED. Never expand boundaries yourself.
- Tests: follow loop pack or reviewer policy; do not invent new test requirements.
- No large logs/diffs in chat; link to
aidd/reports/**. - Block policy:
strictfail-fast by default;strict_recoverable_reason_codesmay explicitly enable bounded retries for selected reasons (for exampleno_tests_hard). ralphuses policy matrix v2 (hard_block|recoverable_retry|warn_continue) with bounded retries for recoverable reasons.- For recoverable blocked retries, emit observability fields:
recoverable_blocked,recovery_path,retry_attempt, plusralph_reason_class|strict_recoverable_reason_class. - Scope mismatch fallback is telemetry-only (
scope_mismatch_non_authoritative=1): emitexpected_scope_key/selected_scope_key; do not treat mismatched fallback as authoritative transition input.
Preload matrix v2
- Roles:
implementer,reviewer,qa. - This skill must not be preloaded for other subagents.
Canonical shared Python entrypoints
python3 ${CLAUDE_PLUGIN_ROOT}/skills/aidd-loop/runtime/loop_pack.pypython3 ${CLAUDE_PLUGIN_ROOT}/skills/aidd-loop/runtime/loop_step.pypython3 ${CLAUDE_PLUGIN_ROOT}/skills/aidd-loop/runtime/loop_run.pypython3 ${CLAUDE_PLUGIN_ROOT}/skills/aidd-loop/runtime/preflight_result_validate.pypython3 ${CLAUDE_PLUGIN_ROOT}/skills/aidd-loop/runtime/output_contract.py- Preflight preparation is stage-chain-internal and not an operator command.
Command contracts
python3 ${CLAUDE_PLUGIN_ROOT}/skills/aidd-loop/runtime/loop_pack.py
- When to run: before loop stage execution to assemble bounded evidence input.
- Inputs: ticket/scope context and available stage artifacts (
readmap,review pack, context pack). - Outputs: deterministic loop pack summary for implement/review/qa orchestration.
- Failure mode: non-zero exit when required upstream artifacts are missing or malformed.
- Next action: repair missing prerequisites and rerun pack generation.
python3 ${CLAUDE_PLUGIN_ROOT}/skills/aidd-loop/runtime/loop_run.py
- When to run: stage-chain-orchestrated loop execution for bounded stage retries.
- Inputs: loop mode/profile, ticket/scope context, and stage-specific payloads.
- Outputs: loop execution status with deterministic retry/blocked metadata.
- Failure mode: non-zero exit on invalid loop payload, policy violation, or unrecoverable blocked condition.
- Next action: apply fix plan or handoff policy action, then rerun with unchanged scope.
Additional resources
- Loop stage-chain reference: reference.md (when: stage-chain path/fallback behavior needs clarification; why: confirm canonical chain and bounded recovery policy).
- Loop pack template source: templates/loop-pack.template.md (when: loop pack format is unclear; why: confirm canonical sections seeded by init/runtime).
Related Skills
grinrus/tasks-new
tools
VerifiedTrustedCommunity
Derives or refines tasklist items from PRD and plan artifacts. Use when tasklist stage prepares implementation-ready work items. Do not use when the request is plan authoring in `plan-new` or loop execution in `implement`.
5SKILL.mdUpdated Apr 5, 2026
grinrus/status
testing
VerifiedTrustedCommunity
Generates consolidated ticket status summary and key artifact pointers. Use when checking current stage state or handoff readiness. Do not use when the request is diagnostics/inventory from `aidd-observability` or flow-state mutation via `aidd-flow-state`.
5SKILL.mdUpdated Apr 5, 2026
grinrus/spec-interview
testing
VerifiedTrustedCommunity
Collects missing spec inputs and synchronizes answers into `spec.yaml`. Use when review-spec leaves unresolved specification fields. Do not use when the request is readiness gating in `review-spec` or task derivation in `tasks-new`.
5SKILL.mdUpdated Apr 5, 2026
grinrus/review
testing
VerifiedTrustedCommunity
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`.
5SKILL.mdUpdated Apr 5, 2026