| name | curator-repo-learn |
| description | Learn patterns from a specific GitHub repository. Clones, analyzes code structure, extracts patterns, and populates procedural memory. Use for: targeted learning from known quality repos, quick knowledge acquisition, specific pattern extraction. Triggers: /repo-learn, /curator-repo-learn, 'learn from repo'. |
| argument-hint | <github-url> [--domain <domain>] |
| user-invocable | true |
| context | fork |
| allowed-tools | ["Task","Bash","Read","Write"] |
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
| Criterion | Weight | Score | Rationale |
|---|
| Coordination Need | 25% | 2/10 | Independent operation |
| Specialization Need | 25% | 9/10 | Requires pattern recognition |
| Quality Gate Need | 20% | 6/10 | Moderate validation needed |
| Tool Restriction Need | 15% | 3/10 | Standard tools sufficient |
| Scalability | 15% | 2/10 | Single repo focus |
| Total | 100% | 5.2/10 | Scenario B optimal |
Workflow (Scenario B)
Task(subagent_type="ralph-researcher", prompt="""
Analyze repository ${REPO_URL}:
1. Clone repository (shallow)
2. Detect domain and language
3. Scan for:
- Function/method patterns
- Class/interface definitions
- Configuration patterns
- Error handling patterns
- Testing patterns
4. Extract rules with:
- domain: <detected>
- category: <detected>
- confidence: 0.75-0.95
- source_repo: ${REPO_URL}
- source_file: <path>
5. Update ~/.ralph/procedural/rules.json
6.
Usage
Basic Usage
/repo-learn https://github.com/owner/repo
With Domain Override
/repo-learn https://github.com/nestjs/nest --domain backend
With Language Hint
/repo-learn https://github.com/vercel/next.js --lang typescript
Process Flow
┌─────────────────────────────────────────────────────────────┐
│ REPO-LEARN PIPELINE │
├─────────────────────────────────────────────────────────────┤
│ │
│ 1. VALIDATE URL │
│ ├── Check GitHub format │
│ └── Verify repository exists │
│ │
│ 2. CLONE REPOSITORY │
│ ├── Shallow clone (--depth 1) │
│ └── Store in ~/.ralph/curator/corpus/learning/ │
│ │
│ 3. DETECT METADATA │
│ ├── Domain (backend, frontend, etc.) │
│ ├── Language (typescript, python, etc.) │
│ └── Framework indicators │
│ │
│ 4. SCAN FILES │
│ ├── Source files (*.ts, *.py, etc.) │
│ ├── Configuration files │
│ └── Documentation files │
│ │
│ 5. EXTRACT PATTERNS │
│ ├── Function signatures │
│ ├── Class structures │
│ ├── Import patterns │
│ ├── Error handling patterns │
│ └── Configuration patterns │
│ │
│ 6. CREATE RULES │
│ ├── rule_id: unique identifier │
│ ├── domain: detected or specified │
│ ├── category: sub-domain │
│ ├── confidence: 0.75-0.95 │
│ ├── source_repo: repository URL │
│ ├── source_file: file path │
│ └── behavior: pattern description │
│ │
│ 7. UPDATE PROCEDURAL MEMORY │
│ ├── Backup existing rules.json │
│ ├── Merge new rules (unique_by rule_id) │
│ └── Update manifest │
│ │
│ 8. CREATE MANIFEST (GAP-C01 FIX) │
│ ├── files[]: processed file list │
│ ├── patterns_extracted: count │
│ ├── detected_domain: domain │
│ └── detected_language: language │
│ │
└─────────────────────────────────────────────────────────────┘
Domain Detection (GAP-C02 FIX)
Automatic domain detection from repository content:
backend: api, server, rest, graphql, controller, service
frontend: react, vue, angular, component, hook, state
database: sql, query, schema, migration, orm, prisma
security: auth, jwt, token, encrypt, hash, csrf
testing: test, spec, jest, vitest, mock, coverage
devops: docker, kubernetes, ci, deploy, pipeline
hooks: hook, lifecycle, callback, trigger, event
general: config, util, helper, common, shared
Output Example
{
"repository": "https://github.com/nestjs/nest",
"learned_at": "2026-02-14T22:00:00Z",
"detected_domain": "backend",
"detected_language": "typescript",
"files": [
"packages/core/nest-application.ts",
"packages/core/nest-factory.ts",
"packages/common/services/logger.service.ts"
],
"patterns_extracted": 12,
"rules_added": [
{
"rule_id": "rule-backend-1739566800-abc123",
"domain": "backend",
"category": "backend",
"source_repo": "https://github.com/nestjs/nest",
"source_file": "packages/core/nest-application.ts"
Error Handling
| Error | Recovery |
|---|
| Invalid URL | Show usage, exit |
| Repository not found | Suggest alternatives, exit |
| Clone failure | Retry with different depth |
| No patterns found | Log warning, return empty |
| Rules merge failure | Restore from backup |
Integration with Curator Pipeline
/curator quick --repo owner/repo
/curator full --type backend --lang typescript
/repo-learn https://github.com/owner/repo
Related Skills
/curator - Full pipeline (Scenario C)/smart-fork - Pattern extraction and forking/research - Repository research
Files
| File | Purpose |
|---|
.claude/scripts/curator-learn.sh | Pattern extraction (GAP-C01, GAP-C02 fixed) |
.claude/scripts/curator-ingest.sh | Repository cloning |
.claude/scripts/backfill-domains.sh | Domain backfill for existing rules |
~/.ralph/procedural/rules.json | Procedural memory |
Action Reporting (v2.93.0)
Esta skill genera reportes automáticos completos para trazabilidad:
Reporte Automático
Cuando esta skill completa, se genera automáticamente:
- En la conversación de Claude: Resultados visibles
- En el repositorio:
docs/actions/curator-repo-learn/{timestamp}.md
- Metadatos JSON:
.claude/metadata/actions/curator-repo-learn/{timestamp}.json
Contenido del Reporte
Cada reporte incluye:
- ✅ Summary: Descripción de la tarea ejecutada
- ✅ Execution Details: Duración, iteraciones, archivos modificados
- ✅ Results: Errores encontrados, recomendaciones
- ✅ Next Steps: Próximas acciones sugeridas
Ver Reportes Anteriores
ls -lt docs/actions/curator-repo-learn/
cat $(ls -t docs/actions/curator-repo-learn/*.md | head -1)
grep -l "Status: FAILED" docs/actions/curator-repo-learn/*.md
Generación Manual (Opcional)
source .claude/lib/action-report-lib.sh
start_action_report "curator-repo-learn" "Task description"
complete_action_report "success" "Summary" "Recommendations"
Referencias del Sistema
