원클릭으로
weekly-reflect
Review recent activity, consolidate memory into atomic notes + MOCs, and prune stale entries
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
메뉴
Review recent activity, consolidate memory into atomic notes + MOCs, and prune stale entries
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
Audit every enabled skill's upstream file dependencies for staleness — flags chained skills about to consume yesterday's article or a long-dead topic file
Audit .github/workflows and composite actions with zizmor + actionlint, classify findings against the prior audit, auto-fix Critical/High regressions, and open a PR only when something actually changed.
Weekly API cost report — computes dollar costs from token usage, flags anomalies, forecasts burn, and prescribes concrete optimizations
Weekly partial-correlation of compute economics against a Hyperliquid macro basket — DePIN-token proxy track runs every week (n>180d), sweep-P&L track defers until n≥30 joined days
Score frontier LLMs (Claude, GPT, Gemini, DeepSeek, Grok) on a private compute-markets task corpus, track score deltas across releases, flag public-vs-private divergence
Save a note as one or more atomic notes under memory/notes/ (and optionally Supernotes). Splits bundled inputs into separate atomic files.
| name | Weekly Reflect |
| description | Review recent activity, consolidate memory into atomic notes + MOCs, and prune stale entries |
| var | |
| tags | ["meta"] |
${var} — Area to focus on. If empty, reviews everything.
If ${var} is set, focus the reflection on that specific area.
Today is ${today}. Review the agent's recent activity and maintain long-term memory under the atomic-notes convention:
memory/MEMORY.md — short index, ~50 lines, no claims.memory/topics/*.md — MOCs that link to atomic notes for durable claims and keep time-varying snapshots (weekly numbers, point-in-time tables) inline.memory/notes/${slug}.md — one file per claim, frontmatter (id, created, type, links), body ≤3 sentences.memory/notes/daily/${today}.md — date-indexed pointer list, written by note-taking.Read state. memory/MEMORY.md, last 7 days of memory/logs/, last 7 days of articles/, memory/skill-health/*.json, and memory/issues/INDEX.md.
Consolidate what you learned.
MEMORY.md? Remove them.lesson).Atomicity pass over memory/notes/. For every file under memory/notes/ (excluding memory/notes/daily/):
links: frontmatter lists the slugs of the others. Delete the original bundled file. Update any [[old-slug]] references in MOCs to point at the most specific replacement.RECONCILE_SPLIT old-slug → new-slug-1, new-slug-2, ….Reorganize memory under the atomic-notes convention.
MEMORY.md stays a pure index. Pointers only, no claims.memory/topics/ become MOCs: a short framing paragraph, then bulleted [[slug]] lists grouped by sub-area, with weekly/snapshot tables kept inline. No multi-sentence claims that should be in a note.memory/notes/${slug}.md (same shape as the note-taking skill produces). Add the [[slug]] link to the right topic MOC.Re-regenerate notegraph so the consolidation shows up:
node scripts/notegraph.mjs
Record the delta (nodes, edges, orphans) in the log. Any new orphans → those notes need links and probably belong in a topic MOC.
Prune. Be ruthless. Memory is a living, useful document — not an append-only log.
status: superseded in its frontmatter and add links: [[newer-slug]]. Don't delete history — let the graph show the supersession.Log what you did to memory/logs/${today}.md:
## Reflect
- **Atomic-pass:** ${SPLIT_COUNT} bundled notes split, ${NEW_COUNT} new atomic notes created
- **Topic MOCs updated:** ${TOPIC_LIST}
- **MEMORY.md pruned:** ${PRUNED_COUNT} stale bullets
- **Notegraph delta:** +${NODE_DELTA} nodes, +${EDGE_DELTA} edges, ${ORPHAN_DELTA:+0} orphans
- REFLECT_OK
Notify via ./notify:
memory consolidated — ${today} · +${NODE_DELTA}n / +${EDGE_DELTA}e
If splits or new atomic notes were created, append (${SPLIT_COUNT} splits, ${NEW_COUNT} new).
A note is atomic if its title states a single claim and the body summarizes in ≤3 coherent sentences without "and also" / "additionally". Bundled notes are linking hazards — every [[X]] ends up pointing at a topic dump instead of one idea, and backlinks lose their definitional power.
When in doubt, split. A claim with rich detail can stay one note; a bundled note is permanently a hazard. Splitting is cheap, merging later is fine too. See PR #260 for the validated effect — across three topic atomizations, edges grew 5.3× faster than nodes.
memory/issues/ISS-NNN.md files — already atomic per filename, don't duplicate as notes.soul/) — voice calibration material, not graph content.node scripts/notegraph.mjs runs in-sandbox; if it fails on a sandbox restriction, log REFLECT_NOTEGRAPH_SKIPPED and continue — the notegraph skill regenerates on its own daily cron.