compact-note — Claude Code skill for trimming notes
The compact-note Claude Code skill condenses long notes and logs while keeping the reasoning behind decisions. It runs on requests to shorten a note, log, or memory.
MIT ★ 26 updated 2026-11-16
npx skills add wrg32786/aigent-os --skill compact-note -a claude-codeWhat compact-note does
The skill reads the whole note and sorts entries into recurring rules, decision-critical records, resolved items, and stale context. A standing rule needs a pattern repeated at least 3 times. Records that explain why a decision was made stay in the note.
It then turns the raw log into a shorter archive, keeps unresolved threads, and updates the note’s metadata with the compaction date and size ratio. A closing ledger records what changed and which entries remain uncertain. The skill requires a full read before editing; partial context is not enough.
What it helps with
- condense a growing session log
- extract repeated rules from a vault note
- retain the reasoning behind past decisions
- trim resolved or outdated entries
Example prompts for the agent
- Condense this note and keep the reasons behind decisions.
- Shorten the log and extract recurring rules.
- Read the entire note, then compress it without dropping important entries.
How to install compact-note
1Skills CLI (Vercel)
npx skills add wrg32786/aigent-os --skill compact-note -a claude-code2Manually: put the files in place
~/.claude/skills/compact-note/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/compact-note/
1Skills CLI (Vercel)
npx skills add wrg32786/aigent-os --skill compact-note -a codex2Manually: put the files in place
.agents/skills/compact-note/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add wrg32786/aigent-os --skill compact-note -a cursor2Manually: put the files in place
.cursor/skills/compact-note/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install wrg32786/aigent-os skills/compact-note2Manually: put the files in place
.github/skills/compact-note/SKILL.md1Skills CLI (Vercel)
npx skills add wrg32786/aigent-os --skill compact-note -a gemini-cli2Platform CLI
gemini skills install https://github.com/wrg32786/aigent-os --path skills/compact-noteManually: put the files in place
.devin/skills/compact-note/SKILL.mdThe legacy .windsurf/skills/ path still works
Skills CLI (Vercel)
npx skills add wrg32786/aigent-os --skill compact-note -a opencodeClaude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about compact-note
How do I install compact-note in Claude Code?
Run npx skills add wrg32786/aigent-os --skill compact-note -a claude-code. The skill file is at ~/.claude/skills/compact-note/SKILL.md.
What does compact-note do?
It condenses long notes by extracting recurring rules, retaining decision reasoning, and turning the raw log into a shorter archive. It also appends a change ledger.
How does compact-note decide which rules to extract?
A rule must come from a pattern repeated at least 3 times. Records that explain decisions are preserved, and uncertain classifications are listed in the ledger.
Source text
SKILL.md · MIT
--- name: compact-note agent: mnemosyne description: "Compact vault notes into standing rules and a shortened log. Use when a note is bloated, a log needs trimming, or memory needs distilling." allowed-tools: Read, Write, Edit, Grep, Glob user-invocable: true triggers: - compact this note - compact note - this note is too long - compress the log - distill this note - trim the vault note - note is getting bloated - needs compact - compact the memory --- # Compact Note You are Mnemosyne compacting a bloated vault note. Distill signal into standing rules. Truncate raw log. Never compact so aggressively that rationale is lost. ## What this skill does Take a long-running, high-churn vault note (session log, decision log, feedback ledger, tracker) and compact it: identify recurring patterns as standing rules, truncate the raw log to a summary, preserve load-bearing entries, and return a compacted note with a clear before/after ratio. ## Protocol ### Step 1: Read the full note Read the entire note. Do not compact from partial context. Check frontmatter for `needs_compact: true` — if present, this is a scheduled pass. ### Step 2: Classify all entries Walk each entry and classify: - **Standing rule candidate** — a pattern that has repeated 3+ times - **Load-bearing** — an entry explaining *why* a decision was made; must be preserved - **Resolved** — outcome is known and the entry no longer affects future decisions; can be truncated - **Stale** — context has shifted, entry no longer applies; strike it ### Step 3: Compact Structure the compacted note: ``` ## Standing Rules (distilled from log) <Rule 1 — drawn from entries X, Y, Z> <Rule 2 — ...> ## Active Threads (load-bearing, not yet resolved) <Thread A — last updated YYYY-MM-DD> ## Archive (truncated log — full history in git) <Summarized in 1-2 bullets per month or sprint> YYYYMM: <N entries> — <one-line summary of what happened> ``` Update frontmatter: - Set `needs_compact: false` - Add `last_compacted: YYYY-MM-DD` - Add `compaction_ratio: X:1` (original lines / compacted lines) ### Step 4: Return honesty ledger ``` ## Honesty Ledger **Notes touched:** <list> **Entries superseded:** <count> **Entries preserved:** <count — with reason for each load-bearing entry> **Compaction ratio:** X:1 (original N lines → compacted N lines) **Residual uncertainty:** <entries that were ambiguous to classify> ``` ## Constraints - Read the full note before compacting. No partial-context edits. - Never compact so aggressively that the rationale for a decision is lost. - Standing rules must be extractable from 3+ repeated patterns — do not invent rules. - Load-bearing entries (the "why" behind decisions) survive compaction. - If in doubt about an entry's load-bearing status, preserve it and flag it.
© wrg32786 · MIT · wrg32786/aigent-os