agent toolingby majiayu000
Curator Repo-Learn Skill (v2.88)
majiayu000/curator-repo-learn
Learn patterns from a specific GitHub repository. Clones, analyzes code structure, extracts patterns, and populates procedural memory.
What it changes
Passed 17 of 20 test cases with the skill, and 7 of 20 without it.
Without skill35%
With skill83%
Checks it passed that the model alone missed
- Includes field files in the manifest output.
- Includes mandatory section Summary in the action report.
- Restores rules.json from backup upon rules merge failure.
and 10 more in the full report
11 fixed8 passed either way4 still failing0 broke
- Pass-rate lift
- +48 pp
- Tokens
- +74%
- Agent turns
- 0%
- Response time
- -68%
Model gemini-3.6-flashMeasured 8/19/2026v119 of 23 cases passed with skill · 20 comparable
See it in action
One of 11 test cases the skill fixed. Same model, same prompt, with and without the skill.
PromptWhat mandatory top-level headings must be included in the generated action report document?
What was checked, without and with the skill| Checked | Without skill | With skill |
|---|
| Includes mandatory section Summary in the action report. | Missed | Passed |
|---|
| Includes mandatory section Execution Details in the action report. | Missed | Passed |
|---|
What's inside
The instructions your model reads when the skill loads.
SKILL.md268 linessingle file
Curator Repo-Learn Skill (v2.88)
Single Repository Pattern Extraction - Learn from a specific GitHub repository.
Role & Priorities
Priorities (ordered): accuracy → relevance → speed → coverage
Scope: Clone repository, analyze structure, extract patterns, update procedural memory.
Agent Teams Integration (v2.88)
Optimal Scenario: B (Custom Subagents)
Why Scenario B for Repo-Learn
- Independent operation: Single repo, single task
- High specialization need: Pattern recognition requires expertise
- Low coordination need: No multi-stage pipeline
- Efficient execution: Direct spawn, no team overhead
Scenario Analysis
Workflow (Scenario B)
# Direct Subagent Spawn
Task(subagent_type="ralph-researcher", prompt="""
Analyze repository ${REPO_URL}:
1. Clone repository (shallow)
Track record
Benchmark
17 / 20
test cases passed, vs 7 / 20 without
Adoption
- Users
- 0
- Times used
- 0
Safety and source details
SkillSafety
✓ 3 of 3 passed✓Code scan15 of 15 checks clean
0 critical · 0 warning · 0 info — scanner v4 — 8/3/2026
View 15 scanner checks
live_secretremote_code_executionreverse_shelldata_exfiltrationssrf_metadataagent_snoopinganti_refusaldestructive_commandsobfuscated_unicodeunicode_homoglyphopaque_blobsuspicious_urltool_overreachtrigger_abuseprompt_injection_phrasing
✓IntentSafe
LLM judge · does the skill do what it claims
✓ContentClean
LLM judge · policy & harmful-content review
Source
Importedmajiayu000 / claude-skill-registry
skills/agent/curator-repo-learn/SKILL.md
License MITImported 8/3/2026Version v1