graduate-insights
Skill109 starsChanged 40 days agoMITView on GitHub
Review and graduate notes to permanent status using Zettelkasten principles. Consolidates AI extractions and document insights into curated permanent notes.
In this repository61 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- adviseSkill
- ai-crystallizeSkill
- analyze-kbSkill
- auto-discoverySkill
- benchmark-memorySkill
- brain-mergeSkill
- changelog-protocolSkill
- coherence-sweepSkill
- compute-lifecycleSkill
- create-articleSkill
- create-explanatory-imageSkill
- decideSkill
- deep-researchSkill
- detect-tensionsSkill
- dialecticSkill
- domain-watchSkill
- elicitation-techniquesSkill
- epistemic-classificationSkill
- epub-chapter-extractorSkill
- extract-document-insightsSkill
- extract-insightsSkill
- find-connectionsSkill
- first-runSkill
- get-perspective-onSkill
- get-youtube-transcriptSkill
- git-commit-pushSkill
- incubation-loopSkill
- ingest-sourceSkill
- insight-capture-formatSkill
- insight-interviewSkill
- integrate-recent-notesSkill
- learn-new-thingsSkill
- manage-reference-dataSkill
- manage-thinking-topicsSkill
- manage-watching-domainsSkill
- multi-format-book-extractorSkill
- nano-banana-image-generatorSkill
- propagate-changeSkill
- quick-searchSkill
- reasoning-checksSkill
- recallSkill
- ref-auditSkill
- ref-bridgeSkill
- ref-ingestSkill
- ref-querySkill
- ref-reconcileSkill
- ref-refreshSkill
- refresh-indexSkill
- ref-supersedeSkill
- scheduled-runSkill
- scope-mountSkill
- search-vaultSkill
- self-diagnosticSkill
- start-hereSkill
- synthesize-insightsSkill
- talkSkill
- test-memory-systemSkill
- think-about-itSkill
- update-changelogSkill
- The first 60 shown.
The file
We split this skill into the rules it gives an agent
12 rules. About 0 are common, 11 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈11
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 85
Delete: - Remove entirely (rare, for duplicates/errors)
- Distinctive: line 89
When user chooses Promote:
- Distinctive: line 112
Verify atomic insight format (single clear idea)
- Very specific: line 113
Ensure wiki-links use [[Note Title]] format
- Distinctive: line 114
Remove source-specific metadata (session info, extraction date)
- Distinctive: line 118
Ask user: "Refine title/content before promoting?
- Distinctive: line 137
If confirmed, remove file.
- Distinctive: line 141
After processing all candidates:
- Distinctive: line 179
When reviewing candidates, consider these principles:
- Distinctive: line 192
Unripe: - Not yet clear how it connects or matters
- Distinctive: line 208
No automated promotion: - Every graduation requires explicit approval
- Distinctive: line 215
After completing this skill's primary task, consider tactical improvements:
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 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
Graduate Insights
Review candidate notes from holding areas and promote valuable ones to permanent note status following Zettelkasten principles.
Purpose
The knowledge base accumulates insights in multiple holding areas:
- AI Extracted Notes/ - Personal insights from conversations (177 notes)
- Document Insights/ - Research findings from external sources (1,608 notes across 89 sessions)
- Books/ - Per-book scopes (the gems mined from each ingested book; the
_book.mdhub is a navigation note, not a candidate) - 00-Inbox/ - Raw captures and staging notes (27 notes)
Without periodic graduation, these insights remain unintegrated. This skill provides a structured review process to:
- Surface high-value candidates based on retrieval frequency and age
- Present each for human judgment (the essential step)
- Graduate worthy notes to
02-Permanent/with proper formatting - Maintain knowledge graph integrity
State Dependencies
| Source | Location | Read | Write |
| Q-values | resources/local-brain-search/data/q_values.json | Yes | No |
| AI Extracted | Brain/AI Extracted Notes/ | Yes | Move |
| Document Insights | Brain/Document Insights/*/ | Yes | Move |
| Book scopes | Brain/Books/*/ | Yes | Move |
| Inbox | Brain/00-Inbox/ | Yes | Move |
| Permanent Notes | Brain/02-Permanent/ | No | Yes |
| Changelog | Brain/CHANGELOG.md | Yes | Yes |
Process
Step 1: Load State and Find Candidates
# Get Q-values for retrieval frequency signals
cat resources/local-brain-search/data/q_values.jsonFind candidate notes prioritized by:
- Q-value > 0 - Notes that have been retrieved and used
- Age > 14 days - Notes that have had time to prove value
- Has connections - Notes with wiki-links to permanent notes
Collect up to 5 candidates per session from:
Brain/AI Extracted Notes/*.mdBrain/Document Insights/**/*.md(exclude CHANGELOG files)Brain/Books/**/*.md(exclude_book.mdhubs and CHANGELOG files)Brain/00-Inbox/*.md
Step 2: For Each Candidate
Present the following information:
## Candidate [N/5]: [Note Title]
**Source:** AI Extracted Notes | Document Insights ([Session Name]) | Inbox
**Created:** [date from frontmatter or file mtime]
**Age:** [X days]
**Q-Value:** [value or "not tracked"]
### Content Preview
[First 500 characters of note content]
### Existing Connections
[List wiki-links found in the note, indicate which link to permanent notes]
### Semantic Neighbors (if available)
[Run quick search to show 3 most similar permanent notes]Ask user:
Decision for "[Note Title]":
- Promote - Graduate to permanent notes
- Skip - Review again later (note stays in place)
- Delete - Remove entirely (rare, for duplicates/errors)
Step 3: Handle Promotion
When user chooses Promote:
- Check for duplicates in
02-Permanent/:# Search for similar titles grep -r "similar keywords" Brain/02-Permanent/ - Prepare note for graduation:
- Ensure frontmatter includes:
--- created: [original date] updated: [today] created_by: [original model if present] updated_by: [current model] agent_version: 01.25 provenance: [see rule below] graduated_from: [original path] graduated_date: [today] --- - Provenance rule (mandatory): default by source -
Document Insights/andBooks/->encountered;AI Extracted Notes/or agent synthesis ->ai-inferred. Preserve an existingprovenance:value if the candidate already carries one. Per CLAUDE.md's guarded boundary, only setoriginatedorendorsedwhen the user explicitly says so at the Promote decision - graduation must never silently mintoriginated. - Verify atomic insight format (single clear idea)
- Ensure wiki-links use
[[Note Title]]format - Remove source-specific metadata (session info, extraction date)
- Note: BDG will auto-classify the layer on next bootstrap (insight or framework based on heuristics)
- Ensure frontmatter includes:
- Optionally refine: Ask user: "Refine title/content before promoting? (or press Enter to keep as-is)"
- Move to permanent:
- Generate clean filename:
Brain/02-Permanent/[Title].md - Move file (not copy - maintains git history)
- Verify move succeeded
- Generate clean filename:
Step 4: Handle Skip
Note remains in original location. No action needed.
Step 5: Handle Delete
Confirm before deleting:
Are you sure you want to DELETE "[Note Title]"?
This cannot be undone. Type 'yes' to confirm.If confirmed, remove file.
Step 6: Session Summary
After processing all candidates:
## Graduation Session Complete
**Reviewed:** 5 notes
**Promoted:** X notes
**Skipped:** Y notes
**Deleted:** Z notes
### Promoted Notes:
- [[New Permanent Note 1]] (from AI Extracted Notes)
- [[New Permanent Note 2]] (from Document Insights/Session Name)
### Remaining Candidates
- AI Extracted Notes: [count] notes
- Document Insights: [count] notes
- Inbox: [count] notes
Run `/graduate-insights` again to continue review.Step 7: Update Changelog
Append to Brain/CHANGELOG.md:
## [DATE] - Insight Graduation Session
**Promoted to Permanent:**
- [[Note 1]] - graduated from AI Extracted Notes
- [[Note 2]] - graduated from Document Insights/[Session]
**Session Stats:** Reviewed 5, Promoted X, Skipped YZettelkasten Graduation Criteria
When reviewing candidates, consider these principles:
Promote If:
- Atomic - Contains ONE clear idea (not a list or summary)
- Evergreen - Will remain true/relevant over time
- Connected - Links meaningfully to existing permanent notes
- Original - Adds unique perspective not already captured
- Actionable - Can inform thinking or decisions
Skip If:
- Ephemeral - Time-sensitive or will become outdated
- Compound - Contains multiple ideas (may need splitting)
- Duplicate - Similar insight exists in permanent notes
- Unripe - Not yet clear how it connects or matters
Delete If:
- Exact duplicate - Same content exists elsewhere
- Error - Extraction mistake or corrupted content
- Obsolete - Information proven wrong or superseded
Outputs
- Promoted notes in
Brain/02-Permanent/with graduation metadata - Updated
Brain/CHANGELOG.mdwith session record - Session summary showing graduation statistics
Notes
- Human judgment is essential - This skill surfaces candidates, humans decide
- No automated promotion - Every graduation requires explicit approval
- Batch size is 5 - Keeps sessions focused and completable
- Priority by Q-value - Notes that have been retrieved rank higher
- Recommend weekly cadence - Prevents backlog accumulation
Self-Improvement
After completing this skill's primary task, consider tactical improvements:
- Review execution: Were there friction points, unclear steps, or inefficiencies?
- Identify improvements: Could error handling, step ordering, or instructions be clearer?
- Scope check: Only tactical/execution changes - NOT changes to core purpose or goals
- Apply improvement (if identified):
- Edit this SKILL.md with the specific improvement
- Keep changes minimal and focused
- Version control (if in a git repository):
- Stage:
git add .claude/skills/graduate-insights/SKILL.md - Commit:
git commit -m "refactor(graduate-insights): <brief improvement description>"
- Stage:
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.

