Pwrl learnings structure

pwrl-learnings-structure — Learning Structuring

Purpose: Phase 3 of learnings workflow. Transforms classified learnings into standardized storage format with comprehensive metadata, full-text indexing, and navigation indexes. Enables efficient retrieval and discovery.

Interaction Method

• Primarily automated transformation and formatting.
• Ask user only for storage strategy preference: "Organize by type or domain?"
• Show storage summary before proceeding.
• No approval gate; proceed to next skill.

Input: Classification Artifact

Expects artifact from pwrl-learnings-classify with:

classify_id: YYYY-MM-DD-NNN-classify
learnings: [array of classified learnings]

Each learning has: type, domain, priority, applicability, tags, title, problem, application.

Output: Structure Artifact

Emit structure artifact (YAML + markdown):

---
format: pwrl-learnings-structure-artifact
version: "1.0"
structure_id: YYYY-MM-DD-NNN-structure
created: ISO-8601-timestamp
---

# Learning Structure Results

## Summary
- **Learnings Stored:** [count]
- **Storage Path:** docs/learnings/
- **Organization Strategy:** by_type | by_domain | by_category

## Storage Structure

### By Type

docs/learnings/ ├── gotcha/ [N files] ├── pattern/ [N files] ├── decision/ [N files] ├── technical_fix/ [N files] └── workflow/ [N files]

### By Domain

docs/learnings/ ├── backend/ [N files] ├── frontend/ [N files] ├── architecture/ [N files] ├── security/ [N files] ├── devops/ [N files] └── ...

## Index Files Generated
- `INDEX.md` — Master index by type
- `BY_TYPE.md` — Organized by type
- `BY_DOMAIN.md` — Organized by domain
- `BY_PRIORITY.md` — Organized by priority
- `BY_APPLICABILITY.md` — Organized by applicability score
- `RECENT.md` — Recently added learnings
- `.index.json` — Machine-readable index

## Storage Details
- **Total Learnings:** [count]
- **Files Created:** [count]
- **Index Files:** [count]
- **Full-Text Index Entries:** [count]

## Format & Metadata
- **Storage Format:** markdown
- **Metadata Fields:** [list of fields per learning]
- **Slug Generation:** enabled (for URLs)
- **Full-Text Index:** enabled (for search)

## Ready for Deduplication
- **Status:** ready
- **Next Skill:** pwrl-learnings-dedup
- **Artifacts Passed:** This structure artifact + learning files

Artifact passed to pwrl-learnings-dedup.

Detailed Workflow

For complete step-by-step instructions, see structure-learnings-detailed-workflow.md.

This SKILL.md provides an overview. The detailed workflow document contains:

• Format normalization with default values
• Slug generation and collision detection
• Fingerprint calculation (for Phase 4 dedup)
• Full-text index entry creation
• Storage path generation by strategy
• Markdown template formatting
• Index file generation examples

Quality Gate Validation

After completing this phase, run quality gate validation:

/pwrl-phase-checkpoint learnings 3 [artifact-path]

See pwrl-phase-checkpoint for validation rules.

---

1. Check that input artifact has:

   • Valid classify_id
   • learnings array populated with classified items
   • Each learning has: type, domain, priority, applicability, tags

2. If verification fails:

   • Return error: "Classification artifact invalid. Return to pwrl-learnings-classify."

Step 2: Determine Storage Strategy

Ask user for preference (if needed):

How should learnings be organized?
- By Type: Gotchas, Patterns, Decisions, etc.
- By Domain: Backend, Frontend, Security, etc.
- By Category: By date (2026-06, 2026-07, etc.)

Default: By Type (most intuitive for browsing)

Step 3: Normalize Learning Format

Ensure all learnings have standard structure:

Learning Record:
  id: [Auto-generated UUID or slug]
  slug: [URL-safe identifier]
  type: [gotcha | pattern | decision | technical_fix | workflow]
  domain:
    [
      backend | frontend | architecture | devops | security | performance | process | testing,
    ]
  priority: [critical | important | nice_to_know]
  applicability:
    current_project: [0-10]
    general: [0-10]
  tags: [list of tags]

  title: [Learning title]
  problem: [Problem description]
  application: [How to apply]

  source:
    type: [code | commit | task | documentation | error | review]
    reference: [file:line | commit_hash | task_id]
    extracted_at: [ISO-8601 timestamp]

  metadata:
    created_at: [ISO-8601 timestamp]
    updated_at: [ISO-8601 timestamp]
    author: [User ID if known]
    confidence: [high | medium | low]

Add defaults for missing fields:

• If id missing: Generate from slug
• If slug missing: Generate from title
• If applicability missing: Auto-calculate from priority/domain

Step 4: Generate Metadata

For each learning, create/enhance metadata:

1. Generate slug:

   • URL-safe, lowercase, hyphenated
   • Example: "race-condition-in-cache"
   • Collision detection: append -1, -2 if needed

2. Generate fingerprints:

   • Title hash (detect duplicates)
   • Content hash (detect modifications)
   • These enable dedup phase

3. Generate full-text index entry:

   • Extract searchable keywords: title + problem + tags
   • Create inverted index entry: keyword → learning ID

4. Add relationships:

   • References to complementary learnings
   • Backreferences from related learnings

Step 5: Generate Storage Paths

Based on strategy, determine file path:

If by_type:

docs/learnings/TYPE/YYYY-MM-DD-SLUG.md
docs/learnings/gotcha/2026-06-12-race-condition-cache.md

If by_domain:

docs/learnings/DOMAIN/YYYY-MM-DD-SLUG.md
docs/learnings/backend/2026-06-12-race-condition-cache.md

If by_category:

docs/learnings/YYYY-MM/TYPE-SLUG.md
docs/learnings/2026-06/gotcha-race-condition-cache.md

Step 6: Format Content

Convert each learning to markdown format:

---
id: [UUID]
type: gotcha
domain: backend
priority: critical
applicability:
  current_project: 9
  general: 8
tags: [typescript, concurrency, cache, critical]
created_at: 2026-06-12T10:30:00Z
source:
  type: code
  reference: "src/cache.ts:42"
---

# Race Condition in Shared Cache

## Problem

Multiple threads can access and modify the cache simultaneously without locking,
causing data corruption and inconsistent state. This can happen when:

- Thread A reads cache value
- Thread B writes new value
- Thread A overwrites with stale value

## Symptom

- Intermittent data corruption
- Tests pass in isolation, fail under concurrency
- Non-deterministic failures (hard to debug)

## Root Cause

Cache object (`this.data`) is mutated without synchronization. JavaScript is
single-threaded, but in Node.js clusters or with workers, multiple processes
can share memory.

## Solution

Use a mutex or semaphore to serialize access:

\`\`\`typescript
import { Mutex } from 'async-lock';

class Cache {
private data = {};
private lock = new Mutex();

async get(key: string) {
return this.lock.acquire('cache', () => this.data[key]);
}

async set(key: string, value: any) {
return this.lock.acquire('cache', () => {
this.data[key] = value;
});
}
}
\`\`\`

## Prevention

1. Always use locking for shared mutable state
2. Prefer immutable data structures
3. Consider using a proper cache library (Redis, Memcached)
4. Test with concurrent operations (stress testing)
5. Use TypeScript strict mode + concurrency linter rules

## Related Learnings

- [Mutex Usage Pattern](mutex-usage-pattern.md)
- [Immutable Data Structure Benefits](immutable-structures.md)

## References

- https://nodejs.org/en/docs/guides/worker-threads/
- MDN: Race Condition

Step 7: Create Index Files

Generate navigation indexes in docs/learnings/ root:

INDEX.md - Master Index:

# Learnings Index

Browse learnings by type:

- [Gotchas](BY_TYPE.md#gotchas)
- [Patterns](BY_TYPE.md#patterns)
- [Decisions](BY_TYPE.md#decisions)
- [Technical Fixes](BY_TYPE.md#technical-fixes)
- [Workflows](BY_TYPE.md#workflows)

Or browse by:

- [Domain](BY_DOMAIN.md)
- [Priority](BY_PRIORITY.md)
- [Applicability](BY_APPLICABILITY.md)
- [Recently Added](RECENT.md)

Total learnings: [count]

BY_TYPE.md - Type Organization:

# Learnings by Type

## Gotchas (20 learnings)

- [Race condition in shared cache](gotcha/2026-06-12-race-condition-cache.md)
- [Null pointer dereference](gotcha/2026-06-10-null-pointer.md)
  ...

## Patterns (30 learnings)

- [Memoization technique](pattern/2026-06-11-memoization.md)
- [Builder pattern for configuration](pattern/2026-06-09-builder-pattern.md)
  ...

BY_DOMAIN.md - Domain Organization:

# Learnings by Domain

## Backend (40 learnings)

- Race Condition (Gotcha) - [link]
- API Rate Limiting (Pattern) - [link]
  ...

## Frontend (20 learnings)

- React Hook Dependencies (Gotcha) - [link]
- Component Composition (Pattern) - [link]
  ...

## Security (15 learnings)

...

BY_PRIORITY.md - Priority Organization:

# Learnings by Priority

## Critical (8 learnings)

- Must know these before shipping
- [link] Race Condition
- [link] SQL Injection Prevention
  ...

## Important (25 learnings)

...

## Nice to Know (20 learnings)

...

BY_APPLICABILITY.md - Relevance Scores:

# Learnings by Applicability

## Highly Relevant (9-10) - 12 learnings

- Current project uses these technologies frequently
- [link] React Hooks
- [link] TypeScript Generics
  ...

## Moderately Relevant (5-8) - 20 learnings

...

## Niche (0-4) - 5 learnings

...

RECENT.md - Recently Added:

# Recently Added Learnings

## Last 30 days (8 new learnings)

- 2026-06-12: Race Condition in Shared Cache (Gotcha)
- 2026-06-11: Memoization Technique (Pattern)
- 2026-06-10: React Custom Hook Pattern (Pattern)
  ...

## Last 90 days

...

Step 8: Generate Full-Text Index

Create .index.json for programmatic search:

{
  "version": "1.0",
  "generated_at": "2026-06-12T10:30:00Z",
  "total_learnings": 80,

  "by_type": {
    "gotcha": 20,
    "pattern": 30,
    "decision": 15,
    "technical_fix": 10,
    "workflow": 5
  },

  "by_domain": {
    "backend": 40,
    "frontend": 20,
    "security": 15,
    "devops": 5
  },

  "learnings": [
    {
      "id": "uuid-1",
      "slug": "race-condition-cache",
      "type": "gotcha",
      "domain": "backend",
      "title": "Race Condition in Shared Cache",
      "priority": "critical",
      "applicability": { "current_project": 9, "general": 8 },
      "tags": ["typescript", "concurrency", "cache"],
      "keywords": ["race", "condition", "cache", "mutex", "lock"],
      "created_at": "2026-06-12T10:30:00Z",
      "path": "gotcha/2026-06-12-race-condition-cache.md"
    },
    ...
  ],

  "keywords": {
    "race": ["uuid-1", "uuid-5"],
    "cache": ["uuid-1", "uuid-3", "uuid-7"],
    "concurrency": ["uuid-1", "uuid-2"],
    ...
  }
}

Step 9: Generate Structure Artifact

Emit artifact with:

• All learnings stored with metadata
• Storage paths and organization strategy
• Index files list
• Full-text index reference
• Navigation structure summary
• Ready flag for next skill

Storage Organization Examples

Example: By Type Organization

docs/learnings/
├── INDEX.md
├── BY_TYPE.md
├── BY_DOMAIN.md
├── BY_PRIORITY.md
├── BY_APPLICABILITY.md
├── RECENT.md
├── .index.json
│
├── gotcha/
│   ├── 2026-06-12-race-condition-cache.md
│   ├── 2026-06-10-null-pointer-dereference.md
│   └── 2026-06-08-async-callback-timing.md
│
├── pattern/
│   ├── 2026-06-11-memoization-technique.md
│   ├── 2026-06-09-builder-pattern.md
│   └── 2026-06-07-middleware-chain.md
│
└── ...

Error Handling

Error	Recovery
Classification artifact invalid	Return error; direct to pwrl-learnings-classify
Storage directory missing	Create docs/learnings/ automatically
Duplicate slug	Append -1, -2, etc. to slug; regenerate path
Slug generation fails	Use UUID as fallback
Index generation fails	Create partial indexes; flag for manual review