info-defrag
Skill16 starsChanged 6 months agoUnlicenseView on GitHub
Use periodically (weekly or before a release) to find stale numbers, outdated descriptions, broken links, and inconsistencies across the codebase. Run after merging multiple PRs or before preparing a meeting report.
In this repository9 other files it gives its agents.
- AGENTS.mdAGENTS.md
- .codex/AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- anti-slop-guideSkill
- prepare-meetingSkill
- run-experimentSkill
- sutro-contextSkill
- sutro-syncSkill
- weekly-catchupSkill
The file
We split this skill into the rules it gives an agent
7 rules. About 0 are common, 7 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈7
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 46
If someone submitted experiments, their bio should mention it.
- Distinctive: line 60
If the gap is more than a week, extend the timeline.
- Distinctive: line 64
Check that index files list all their children.
- Distinctive: line 82
Check that files mentioned in docs actually exist.
- Distinctive: line 94
Check if TODO.md has items marked done that aren't actually done, or open items that were completed.
- Distinctive: line 107
Before preparing a weekly catch-up
- Distinctive: line 109
When someone says "is this up to date?"
≈ “Common” is an estimate: at least 1 in 5 of the 372,429 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
Info Defrag
Scan the codebase for stale information, inconsistencies, and broken references. This catches drift that happens when experiments get merged but docs don't get updated.
What to check
1. Experiment count
The number of experiments appears in multiple files. They drift.
# Actual count
wc -l research/log.jsonl
grep -c "^| exp" DISCOVERIES.md
# Where the number appears
grep -rn "33 experiment\|34 experiment\|35 experiment\|36 experiment" docs/ CLAUDE.md README.mdFix any file that shows a different number than the log.
2. Best methods table
CLAUDE.md and DISCOVERIES.md both have a methods ranking table. They should match.
# Compare the two
grep -A10 "Current Best Methods" CLAUDE.md
grep -A10 "DMC baseline rankings" DISCOVERIES.md3. People descriptions
Contributors change roles. Check docs/context.md People table against recent git log.
# Who contributed recently
git log --format="%an" --since="2 weeks ago" | sort -uIf someone submitted experiments, their bio should mention it.
4. Timeline coverage
The gantt chart in docs/context.md should cover recent work.
# Last date in the timeline
grep -o "2026-[0-9-]*" docs/context.md | sort | tail -1
# Last commit date
git log --format="%ai" -1 | cut -d' ' -f1If the gap is more than a week, extend the timeline.
5. Index pages
Check that index files list all their children.
# Catchups
ls docs/catchups/*.md | wc -l
grep -c "\.md" docs/catchups/index.md
# Sessions
ls docs/sessions/*.md | grep -v transcript | wc -l
grep -c "\.md" docs/sessions/index.md
# Findings in nav
ls docs/findings/exp*.md | wc -l
grep -c "findings/" mkdocs.yml6. Broken file references
Check that files mentioned in docs actually exist.
# Find markdown links and check targets
grep -roh '\[.*\](\.\./[^)]*\.md)' docs/ | grep -o '(.*\.md)' | tr -d '()' | sort -u | while read f; do
resolved="docs/$f"
[ ! -f "$resolved" ] && echo "BROKEN: $f"
done7. Stale TODO items
Check if TODO.md has items marked done that aren't actually done, or open items that were completed.
# Open items
grep -c "^\- \[ \]" TODO.md
# Compare with closed issues
gh issue list --repo cybertronai/SutroYaro --state closed --json title --jq '.[].title' | head -10When to run
- After merging 2+ PRs
- Before preparing a weekly catch-up
- Before a meeting
- When someone says "is this up to date?"
Output
List every stale item with: file, line, what it says, what it should say. Don't fix anything automatically. Report the findings so the user can decide what to update.
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.

