From cortexloop
Shared protocol contract for 7-pass expert pipeline: enforces single-category analysis, handoff JSON output, and defer boundaries. Load this on every pipeline expert before domain skills.
How this skill is triggered — by the user, by Claude, or both
Slash command
/cortexloop:cortexloop-expert-coreThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Ultra-thin shared contract for all 7 pipeline experts. **Load this first**, then your pass contract (`passes/XX-*.md`) and domain depth skill(s) only.
Ultra-thin shared contract for all 7 pipeline experts. Load this first, then your pass contract (passes/XX-*.md) and domain depth skill(s) only.
deferToLaterPasses with target pass key + one-line note.Diff wins over prior report claims when they disagree.
A scored finding requires:
Low-confidence ideas → openQuestions only, not findings.
### [SEVERITY] [Title]
- **Location:** path:line
- **Category:** <your pass category only>
- **Problem:** ...
- **Evidence:** ...
- **Confidence:** high | medium
- **Recommendation:** ...
- **Auto-fixable:** yes | no | needs-confirmation
Map severity to: Critical / High / Medium / Low / Info.
Write to the path in your pass contract. Required fields:
pass, category, expert, summary (1–3 sentences for downstream experts)findings — scored items onlydeferToLaterPasses — [{ "pass": "<passKey>", "note": "..." }]openQuestions — unresolved, not scoredSchema: schemas/pass-handoff.schema.json
Also write the category markdown report path from your pass contract.
validate-handoffs.mjs before scoringIf a later pass deferred an issue to your pass (you already ran in Step 3), orchestrator may re-delegate a targeted catch-up. Then:
crossValidation[] on your handoff with orphanId + verified or rejectedverified, append a scored finding in your category and update category markdownnpx claudepluginhub whitequeen306/code-cortex-loop --plugin cortexloopMulti-phase code review pipeline with mechanical checks, graph-scoped context, parallel review agents, cross-agent deduplication, and structured output. Use for completed work reviews.
Runs a multi-agent validation loop that reviews code, autonomously fixes issues via a subagent, and retries until all agents approve. Use when implementation is complete and you need thorough verification.
Classifies code findings into three axes (bug, scatter, dead/smelly) with mandatory coaching and citation checks. Automatically invoked by radar-suite audit skills before emission.