메뉴
Audit and sync agent brain — read all rules from disk, reconcile skills/catalogs/routing. Triggers on refresh agent brain, refresh all rules, audit rules, or after steering promotion. Read-only audit script; agent writes doc updates.
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
SOC 직업 분류 기준
Runs StepCOVNet GPU training and MERT extraction on Windows via WSL with TensorFlow CUDA libraries. Use for train_onset, overfit scripts, extract_mert_features with cuda, or when TensorFlow sees zero GPUs in WSL.
One-prompt agent autoresearch loop for StepCOVNet: goal + time budget → orient, hypothesize, run, log, repeat until success or deadline. Use when the user says "run autoresearch", gives a research goal with a budget, or names a profile (e.g. ar-tide-overfit). Not for unattended Python planners (--hours / lattice).
Deprecated alias — use autoresearch skill with profile ar-tide-overfit.
Starts a research session and logs EXP and NOTE entries with timestamps. Use when beginning research work, logging experiments, or writing discussion notes. Promote findings to PAPER_OUTLINE only when drafting the paper.
Triggers on process mistakes and steering corrections; run steering-correction-promotion for artifact choice, context optimization, and JRN receipt. Journal alone is incomplete.
On user steering corrections or remember-this, pick the smallest durable artifact (skill section, scoped rule, or rare always-apply rule), slim AGENTS.md, optimize agent-brain context, then prepend JRN. Use instead of creating new alwaysApply rules by default.
| name | agent-brain-refresh |
| description | Audit and sync agent brain — read all rules from disk, reconcile skills/catalogs/routing. Triggers on refresh agent brain, refresh all rules, audit rules, or after steering promotion. Read-only audit script; agent writes doc updates. |
| disable-model-invocation | true |
Holistic pass over rules, skills, and routing docs so they match disk and stay context-efficient.
User phrases refresh all rules, refresh agent brain, and audit rules all invoke this skill (full refresh).
| You do | You do not |
|---|---|
Read every .cursor/rules/*.mdc from disk this turn — re-ground on bodies, not only agent-brain.md or audit stdout | Assume summarized chat memory is current |
Reconcile rules ↔ skills ↔ AGENTS.md ↔ catalogs; optimize scope and dedupe | Reload Cursor’s rule injection (alwaysApply is automatic) |
Run audit_agent_brain.py for inventory + drift; you edit catalogs if needed | Let the script write markdown |
| Report clean — no edits when audit OK and content already agrees | Treat audit-only as sufficient without reading rule files |
| Edit rule/skill/router files only when judgment finds drift, waste, or duplication | Add always-on rules or load multiple AGENTS.md indexes |
Rules are the center; the brain is the scope. “All rules” means each .mdc is read and judged; refresh still covers skills and routing docs that must stay aligned with those rules.
| Trigger | Action |
|---|---|
| refresh all rules, refresh agent brain, audit rules, brain refresh | Full refresh (below) |
| After steering-correction-promotion | Quick refresh (below) |
| Session end (with agent-self-improvement) | Quick refresh if any brain file changed this session |
| Periodic | Every ~3 steering promotions or when audit_agent_brain.py reports drift |
| File | Role |
|---|---|
.cursor/rules/*.mdc | Source of truth for behavior — Cursor loads alwaysApply automatically |
| docs/agents/agent-brain.md | Human catalog — you update during refresh; not script-generated |
| AGENTS.md | Router only — task index; link to catalog |
| .cursor/skills/README.md | Skill index + scoped-rules summary |
scripts/audit_agent_brain.py | Read-only — prints disk inventory + drift; does not write files |
Minimize alwaysApply rules — each one loads every turn. Prefer scoped rules (globs) or skills. No hard budget, but justify new always-on rules in JRN.
Read each file in .cursor/rules/*.mdc (always-apply and scoped). Note scope, overlap with skills, and stale guidance.
venv\Scripts\python.exe scripts/audit_agent_brain.py
Use stdout disk inventory as a checklist. Fix any DRIFT lines.
| File | You edit when |
|---|---|
| agent-brain.md | Rules or skills added/removed/demoted |
| skills README | Playbooks row + scoped-rules table |
| AGENTS.md | New routing index row only |
| Check | Fix |
|---|---|
Each */SKILL.md under .cursor/skills/ | Row in skills README Playbooks |
Each scoped .mdc | Row in skills README Agent brain table + agent-brain.md |
| New skill needs routing | One row in AGENTS.md task index only |
| Procedure text in AGENTS.md | Move to skill; leave link |
alwaysApply: true → set globs, alwaysApply: falsevenv\Scripts\python.exe scripts/audit_agent_brain.py
Exit 0 before finishing.
Prepend JRN with Artifact paths when refresh changed rules, catalog, or indexes.
.mdc / skill filesaudit_agent_brain.py (read)audit_agent_brain.py to write or overwrite markdown