knowledge-sync
Skill17 starsChanged 50 days agoMITView on GitHub
Sync vault knowledge patterns to skills. Scans Second Brain for strong signals, proposes skill updates. Use when user says knowledge sync, sync skills, update skills from vault, or promote patterns.
3 other projects use a version of this.
Adopted by
Most recently taken up by
- pierrebelin/claude-code-toolkittheir own version, first seen Nov 1, 2026
- Isabella-a/Skillstheir own version, first seen Oct 19, 2026
- microsoft/powerplatform-build-toolstheir own version, first seen Apr 8, 2026
Other versions3 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 3
- microsoft/powerplatform-build-tools★ 314same name
- pierrebelin/claude-code-toolkit★ 7same name
- Isabella-a/Skills★ 2same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository37 other files it gives its agents.
- adrSkill
- android-kotlinSkill
- apple-swiftSkill
- autoresearch-promptSkill
- blog-writerSkill
- cloud-infrastructureSkill
- commitSkill
- cover-imageSkill
- freezeSkill
- gemini-reviewSkill
- golangSkill
- harness-mechanicSkill
- harness-traceSkill
- humanizerSkill
- ios-debuggerSkill
- learning-docsSkill
- learning-loopSkill
- obsidianSkill
- orchestratorSkill
- plan-forgeSkill
- process-clippingsSkill
- project-analyzerSkill
- project-checksSkill
- pr-reviewSkill
- pythonSkill
- railsSkill
- react-nextjsSkill
- refine-requirementsSkill
- releasing-softwareSkill
- rubySkill
- scoreSkill
- second-opinionSkill
- skill-forgeSkill
- source-controlSkill
- terraformSkill
- test-design-reviewerSkill
- verify-frontendSkill
The file
We split this skill into the rules it gives an agent
8 rules. About 0 are common, 5 are distinctive, and 3 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈5
Distinctive · estimate
This author's own choices. Worth a look.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 41
FILTER: only strong signals; skip rows where signal = applied or pattern already exists in target skill
- Distinctive: line 49
After human approval for each candidate:
- Very specific: line 52
Add source backlink as HTML comment: <!-- vault: [[Second Brain - Development#Pattern Name]] -->
- Very specific: line 53
Update the Skill Candidates table: change signal from strong to applied
- Distinctive: line 98
Read the target skill file before proposing additions (avoid duplicates)
- Distinctive: line 102
Never auto-apply: Always present proposals and wait for approval.
- Distinctive: line 103
One skill at a time: Don't batch across skills without per-skill approval.
- Distinctive: line 105
Backlinks are mandatory: Every applied pattern traces back to its vault source.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
ABOUTME: Vault-to-skills sync, promotes recurring patterns from Second Brain to skill files
ABOUTME: Scan/filter/group/propose/approve/apply cycle with signal strength tracking
Knowledge Sync
Promotes recurring patterns from vault Second Brain notes into skill files. The APPLY step runs on human approval only, never autonomously. The scan-and-propose phase MAY run on a schedule (see Scheduled Mode).
Signal Strength
| Signal | Criteria | Action |
| weak | 1 project or 1 source | Track, don't propose |
| strong | 3+ projects OR 2+ independent sources | Propose to human |
| applied | Already synced to skill | Skip |
Annotation Schema
Second Brain notes use a ## Skill Candidates table:
## Skill Candidates
<!-- Updated by learning-docs during retrospectives -->
| Pattern | Target Skill | Seen In | First Seen | Signal |
|---------|-------------|---------|------------|--------|
| Go error wrapping with %w | golang | feed-brain, bbp | 2026-01-15 | weak |
| Token bucket rate limiting | cloud-infrastructure | feed-brain, nur, clawdbot | 2026-01-10 | strong |Table-based: easy to append via CLI, readable in Obsidian, parseable by Claude.
Process
- SCAN:
obsidian search query="## Skill Candidates" path="Second Brain" matches - FILTER: only
strongsignals; skip rows where signal =appliedor pattern already exists in target skill - GROUP: candidates by target skill with vault source citations
- PROPOSE: draft exact text to add to each skill file (show diff-style)
- APPROVE: human gate (always RED in decision framework)
- APPLY: edit skill file + add backlink + update signal
Apply Step Details
After human approval for each candidate:
- Edit target skill file with the proposed addition
- Add source backlink as HTML comment:
<!-- vault: [[Second Brain - Development#Pattern Name]] --> - Update the Skill Candidates table: change signal from
strongtoapplied
The Obsidian CLI only appends or prepends, so it cannot change a table row in place. Read the note with obsidian read, then change the row's signal with the Edit tool on the note's file under the vault root (path in ../_OBSIDIAN.md, Path fallback).
Presentation Format
## Knowledge Sync Report
### Candidates Found: N strong, M weak (skipped)
#### golang (2 candidates)
**1. Go error wrapping with %w**
- Source: [[Second Brain - Development#Go Error Handling]]
- Seen in: feed-brain, bbp
- Proposed addition to `skills/golang/SKILL.md`:[exact text to add]
- [ ] Approve [ ] Skip [ ] Modify
#### cloud-infrastructure (1 candidate)
...Scheduled Mode (propose-only)
The scan half of this skill may run unattended; the APPLY gate never does. A scheduled run executes SCAN → FILTER → GROUP → PROPOSE, writes the report to quality_reports/knowledge_sync/report-YYYY-MM-DD.md in the forge repo, and STOPS. Max reviews the report later and triggers APPLY interactively.
Runner: a user crontab entry (1st of the month, 09:23 local). The human installs it manually: persisting a headless autonomous run is a user-authorized action, Claude only documents the line (crontab -l to inspect, crontab -r or edit to remove). Replace <forge-repo> and <claude-bin> with your paths (which claude), and mkdir -p ~/.claude/logs first (cron fails silently on a missing redirect target):
23 9 1 * * cd <forge-repo> && <claude-bin> -p "Run the knowledge-sync skill in Scheduled Mode: SCAN, FILTER, GROUP, PROPOSE only. Write the report to quality_reports/knowledge_sync/report-$(date +\%Y-\%m-\%d).md. Do NOT apply any skill changes." >> ~/.claude/logs/knowledge-sync-cron.log 2>&1Caveats:
- cron does not catch up on runs missed while the machine sleeps: if the report is missing after the 1st, kick the same prompt manually.
- The headless run depends on the settings.json allowlist covering the
obsidianCLI; a permission denial shows up in the log file, not on screen. - The report lands on whatever branch is currently checked out in the forge repo. Reports are tracked in git (review history), not gitignored.
- CronCreate (in-session scheduler) is NOT suitable here: its jobs are session-only and expire within 7 days.
Quality Notes
- Read the target skill file before proposing additions (avoid duplicates)
Rules
- Never auto-apply. Always present proposals and wait for approval.
- One skill at a time. Don't batch across skills without per-skill approval.
- Preserve existing content. Additions only, never modify existing skill text.
- Backlinks are mandatory. Every applied pattern traces back to its vault source.
- Cadence: monthly or after major project milestones. Not after every session.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

