Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add commands/kimsanguine/llm-brain/curategit clone --depth 1 https://github.com/kimsanguine/llm-brainWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/commands/kimsanguine/llm-brain/curate)<a href="https://agentmods.dev/commands/kimsanguine/llm-brain/curate"><img src="https://agentmods.dev/badge/commands/kimsanguine/llm-brain/curate.svg" alt="Measured on agentmods" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00022 | $0.01128 |
| Opus 5 | $0.00011 | $0.00564 |
| Sonnet 5 | $0.00004 | $0.00226 |
| Haiku 4.5 | $0.00002 | $0.00113 |
Grade A, and why
curate scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 5d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 65 lines — stays where its author put it; the contents beside it link to each section on GitHub.
llm-brain의 curate 커맨드입니다. $ARGUMENTS에 따라 아래를 실행하세요.
인자 파싱
--distill→ distill 모드--lifecycle→ lifecycle 모드--audit→ audit 모드--all또는 인자 없음 → 전체 실행 (audit + distill + lifecycle)--purge→ archive 후보 실제 이동--reweave→ reweave 모드 (매일 실행 —run_daily.shStep 4). 조합:--fix(자동 보강 가능분 즉시 수정) ·--dry-run(변경 없이 스캔만) ·--weekly-summary(일요일, 4주 누적 weak 요약)
그래프 탐색은
/ingest(delta canvas)와/query(neighborhood canvas)가 담당합니다.
Step 1: curate 스크립트 실행
cd "$(git rev-parse --show-toplevel)" # llm-brain 레포 루트
uv run python scripts/curate.py $ARGUMENTS
실행 결과로 wiki/curate_report.md, wiki/distill_queue.md(--distill/--all 시), wiki/reweave_queue.md(--reweave 시)가 생성됩니다.
Step 2: distill 실행 (--distill 또는 --all 일 때)
wiki/distill_queue.md의 긴급 후보(access ≥ 10, distill_level < 3)를 순서대로 압축합니다:
각 대상 페이지에 대해:
- 현재 페이지 내용 읽기
distill_level에 따라 압축:- level 0 → 1: 원문에서 핵심 단락만 추출 (50% 축약)
- level 1 → 2: 핵심 개념과 관계만 남기기 (30% 이하)
- level 2 → 3: 한 문장 핵심 요약 추가
- frontmatter의
distill_level값 +1,last_distilled오늘 날짜로 갱신
Step 3: reweave 실행 (--reweave 일 때)
Step 1의 스크립트가 산출한 wiki/reweave_queue.md(weak-node 판단 필요분 ## 판단 필요분 + 종합 대상 ## 종합 대상(v0.3.1 WS-1), distill_queue와 동일 체크박스 패턴)와 curate_report.md의 ## Reweave 섹션 alert(판단 필요분: 본문·근거 부족 — 자동 fix 불가분)를 읽고 처리합니다:
- 보강 (weak-node 판단 필요분): 각 대상 페이지에 대해
- 페이지 본문과 frontmatter
sources의 raw/ 파일 읽기 - raw 근거 범위 안에서만 보강 —
schema/curate.md의 Promotion Gates G-2 기준 충족 (사례 ≥1건 OR 새 각도 ≥200자, 강화 후 본문 ≥800자) - raw 근거가 부족하면 보강하지 않고 alert 그대로 유지 (가짜 보강 금지)
- 페이지 본문과 frontmatter
- 종합 (synthesis 대상):
schema/curate.md의## Synthesis Rules에 따라## 인사이트 (종합)섹션 생성·갱신 + frontmatterangles/signal_count/synthesis_updated갱신 - 처리한 항목은 큐의 체크박스를 체크
- 불변식: 기존 본문·sources 삭제·단축 금지(append/갱신만) · raw 출처 없는 서술 금지
--weekly-summary 시: 4주 누적 weak의 통합/삭제 후보는 목록 보고만 — 실제 이동·삭제는 사용자 승인 필수.
Step 3.5: 모순 후보 화해 (wiki/contradiction_queue.md 존재 시)
--audit/--all 이 모순 후보를 감지하면 wiki/contradiction_queue.md(후보 ≥1일 때만 생성)를 읽고, schema/curate.md의 ## Reconciliation Rules에 따라 각 대상 페이지에 ## 반론/갱신 (YYYY-MM-DD) 3요소(기존 주장·반례 근거·현재 판단)를 append 하고 frontmatter superseded_claims/last_reconciled를 갱신합니다. 옛 주장 삭제 금지(본문 유지 + 표시만) · 오탐이면 반론 미생성 · raw/ 출처 뒷받침 범위까지만 서술.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 5d ago First seen · 65 lines · 22 tokens per session scan A 7f00a29f165f
curate is a command published in the GitHub repository kimsanguine/llm-brain (10 stars, last pushed 21d ago), licensed MIT. It adds 22 tokens to every session and 1,128 once invoked, about $0.0001 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other commands, from other repositories
onboard
Interview-based first-run setup for this LLM Wiki kit. Asks the essential questions (vault location/name, Mode A/B, mothership path, Core Context identity + reuse axes), then fills every placeholder and writes Core Context so the wiki knows you from day one. Activate when the user says "온보딩해줘", "처음 시작할게", "처음 시작"…
refresh-context
Re-snapshot Core Context.md from (optional) mothership system files + key personal essays. Run when mothership files change, when new essays signal philosophy shift, or when snapshot is older than 30 days.
ingest
Ingest a source (URL/file/text) into Raw Sources + compile 1015 Wiki pages, with mandatory user-purpose gate and mothership cross-linking.
audit
Audit the whole Wiki vault against 3 knowledge-integrity criteria — eligibility coverage, MOC-cluster consistency, confidence calibration. Produces a vault health report and queues high-priority pages for /verify.
verify
Verify a single Wiki page against 3 knowledge-integrity criteria — eligibility, consistency, confirmability. Writes verificationStatus back to the page; flags conflicts as disputed rather than deleting them.
lint
Run comprehensive wiki health check — orphans, broken links, contradictions, stale pages, index sync, MOC coverage, v2/v4/v5 frontmatter coverage, Core Context freshness, and cross-vault link integrity (mainVaultRelated/mainVaultCmds).