
proposed-name
by Yehonatan-Bar
Processing documentation into skills
SKILL.md
description: MANDATORY - INVOKE IMMEDIATELY after completing ANY development task (feature, bug fix, refactor, maintenance). Creates SR-PTD documentation for future skill extraction. This is NOT optional - every task MUST have documentation.
SR-PTD Documentation Skill
Purpose
Capture exactly what happened during a task in a structure that maps directly to Claude Skill architecture. Over time, these documents become raw material for building Skills - patterns emerge, reusable components accumulate, and workflows crystallize.
Key Principle: Every task you complete is a potential Skill waiting to be extracted. Document it accordingly.
How SR-PTD Maps to Skills
| Document Section | Skill Component | Extraction Value |
|---|---|---|
| Task Trigger Profile | YAML description | What requests should activate this skill? |
| Workflow Executed | SKILL.md body | Step-by-step process to replicate |
| Knowledge Accessed | references/ | Schemas, APIs, docs Claude needed |
| Code Written/Used | scripts/ | Reusable scripts for future tasks |
| Outputs Produced | assets/ | Templates, formats to standardize |
Process
Step 1: Determine Document Type
Choose based on task complexity:
Full SR-PTD (Sections A-J): For complex tasks, multi-step implementations, tasks likely to become skills Quick Capture: For simpler tasks, quick fixes, routine work
Step 2: Create Documentation File
Filename Format: SR-PTD_[YYYY-MM-DD]_[task-id]_[brief-description].md
Save Location: C:\projects\Skills\Dev_doc_for_skills\
Step 3: Fill All Applicable Sections
Follow the template structure below. Be thorough - this documentation serves as the foundation for future Skills.
Step 4: Assess Skill Potential
Score the task's reusability. High scores (20+) indicate strong skill candidates.
Step 5: If Conversation Continues
Update the existing document instead of creating a new one. One document per session.
Full SR-PTD Template
Section A - Header and Skill Trigger Profile
# SR-PTD - [Task ID or Brief Description]
## Section A - Header and Skill Trigger Profile
### Metadata
- **Date**: YYYY-MM-DD
- **Task ID/Ref**: [ticket number, request ID, or descriptive identifier]
- **Type**: Feature | Bug Fix | Maintenance | System Update | Refactor | Research | Automation
- **Domain/Module**: [e.g., auth, payments, meetings, reports]
- **Complexity**: Low | Medium | High | Critical
- **Time Spent**: Total __h (Planning __h | Execution __h | Verification __h)
### Skill Trigger Profile (Future YAML description)
**What triggered this task?**
> [Exact user request or problem statement that initiated this work]
**Keywords/Phrases that indicated this task:**
> [List the words/concepts that signaled this type of work was needed]
**Context markers:**
- File types involved: [.py, .docx, .xlsx, etc.]
- Systems touched: [DB name, API name, service name]
- Domain concepts: [meetings, reports, users, payments, etc.]
**Draft Skill Trigger** (if this became a skill, when should it activate?):
> [Write 1-2 sentences describing when this skill should trigger]
Section B - Context and Inputs
## Section B - Context and Inputs
### Problem Statement
- **Objective**: [1-2 sentences - what needed to be accomplished]
- **Why it mattered**: [Business value or technical necessity]
- **Success criteria**: [How did we know when it was done?]
### Starting State
- **Files/Data received**:
- `[filename]` - [purpose/contents]
- `[filename]` - [purpose/contents]
- **Existing code/configs touched**:
- `[path/file]` - [what it contained]
- **Relevant tickets/docs**: [Links]
### Environment
- **Runtime**: [Local | CI | Staging | Prod]
- **Tool versions**: [Python 3.11, Node 18, etc.]
- **Dependencies used**: [Key packages]
### Constraints and Dependencies
- **Deadlines**: [if any]
- **External dependencies**: [APIs, services, people]
- **Blockers encountered**: [if any]
Section C - Workflow Executed (Future SKILL.md Body)
## Section C - Workflow Executed
### Workflow Type
- [ ] Sequential (steps in fixed order)
- [ ] Conditional (branching based on input/state)
- [ ] Iterative (repeat until condition met)
- [ ] Hybrid
### High-Level Steps Taken
1. [Step name] - [One-line what was done]
2. [Step name] - [One-line what was done]
3. [Step name] - [One-line what was done]
...
### Detailed Step Log
#### Step 1: [Step Name]
- **Action**: [What was done]
- **Tool/Command**:
```bash
[Exact command or code snippet]
- Input: [What this step received]
- Output: [What this step produced]
- Decision made: [If any choice was made, what and why]
- Time: [Optional]
Step 2: [Step Name]
- Action:
- Tool/Command:
- Input:
- Output:
- Decision made:
[Continue for each meaningful step...]
Decision Points Encountered
| Decision | Options Considered | Choice Made | Rationale |
|---|---|---|---|
| [What needed deciding] | [Option A, Option B] | [Chosen option] | [Why] |
Workflow Diagram (if complex)
START
|
+-> [Step 1]
| |
| v
+-> [Step 2] --> Condition? --> YES -> [Path A]
| |
| +-> NO -> [Path B]
|
+-> [Final Step]
### Section D - Knowledge Accessed (Future references/)
```markdown
## Section D - Knowledge Accessed
### Database/Data Knowledge Used
**Tables Queried:**
| Table | Columns Used | Purpose in This Task |
|-------|--------------|---------------------|
| | | |
**Queries Written:**
```sql
-- [Purpose of query]
[Actual SQL]
Schema Knowledge Required:
[What did you need to know about the data structure?]
Relationships Leveraged:
[What table joins or foreign keys were important?]
API Knowledge Used
Endpoints Called:
| Method | Endpoint | Purpose |
|---|---|---|
Request/Response Patterns:
// Request to [endpoint]
{
}
// Response
{
}
Authentication Used:
[How did you authenticate?]
Rate Limits/Errors Encountered:
[Any API-specific issues?]
Codebase Knowledge Used
Files Read/Modified:
| File Path | Why Accessed | Key Content |
|---|---|---|
Patterns/Conventions Followed:
[What naming conventions, code patterns, or architectural decisions did you follow?]
Dependencies/Imports Required:
# Key imports used
from x import y
Documentation/External Knowledge Used
Docs Consulted:
- [Link or reference] - [What you learned from it]
- [Link or reference] - [What you learned from it]
Domain Knowledge Applied:
[What business logic, rules, or domain concepts were essential?]
### Section E - Code Written/Used (Future scripts/)
```markdown
## Section E - Code Written/Used
### New Code Created
#### Script/Function 1: [Name]
**Purpose**: [One line]
**Reusability**: [ ] One-time | [ ] Likely reusable | [ ] Definitely reusable
```python
# [filename].py
[Full code or key sections]
Inputs: [Parameters, files, etc.] Outputs: [Return values, files created, etc.] Dependencies: [Required packages]
Should this become a skill script? [ ] Yes | [ ] No Why/Why not:
Script/Function 2: [Name]
Purpose: Reusability: [ ] One-time | [ ] Likely reusable | [ ] Definitely reusable
Should this become a skill script? [ ] Yes | [ ] No Why/Why not:
Existing Code Reused
| Code Source | What It Did | Modifications Made |
|---|---|---|
| [path or snippet name] |
Code That Should Be Extracted to Script
| Functionality | Current Location | Extraction Priority | Notes |
|---|---|---|---|
| [What it does] | [Inline/ad-hoc] | High | Medium |
### Section F - Outputs Produced (Future assets/)
```markdown
## Section F - Outputs Produced
### Files Created
| Filename | Format | Purpose | Template Potential |
|----------|--------|---------|-------------------|
| | | | [ ] Could be template |
| | | | [ ] Could be template |
### Output Patterns/Formats Used
**If output followed a structure, document it:**
```markdown
# [Output Type] Structure
## Section 1: [Name]
[What goes here]
## Section 2: [Name]
[What goes here]
Should this become a template asset? [ ] Yes | [ ] No Why/Why not:
Generated Artifacts
- Reports: [Paths/links]
- Data files: [Paths/links]
- Configs created: [Paths/links]
- PRs/Commits: [Links]
### Section G - Issues and Fixes
```markdown
## Section G - Issues and Fixes
### Issues Encountered
#### Issue 1: [Brief Name]
- **Symptom**: [What went wrong / what you observed]
- **Root Cause**: [Why it happened]
- **Fix Applied**: [What you did]
- **Time Lost**: [Optional]
- **Prevention**: [How to avoid in future]
#### Issue 2: [Brief Name]
- **Symptom**:
- **Root Cause**:
- **Fix Applied**:
- **Prevention**:
### Error Patterns Worth Documenting
| Error Type | Cause | Solution | Include in Skill? |
|------------|-------|----------|-------------------|
| | | | [ ] Yes |
| | | | [ ] Yes |
Section H - Verification and Validation
## Section H - Verification and Validation
### Tests Run
| Test Type | What Was Tested | Result | Evidence |
|-----------|-----------------|--------|----------|
| Manual | | Pass/Fail | |
| Unit | | Pass/Fail | |
| Integration | | Pass/Fail | |
| E2E | | Pass/Fail | |
### Validation Evidence
- **Screenshots**: [Links/paths]
- **Logs**: [Links/paths]
- **Metrics**: [Before/after if applicable]
### Success Criteria Verification
| Criterion | Met? | Evidence |
|-----------|------|----------|
| [From Section B] | Yes/No | |
| | | |
Section I - Skill Extraction Assessment
## Section I - Skill Extraction Assessment
### Reusability Score
Rate this task's potential for skill extraction:
| Dimension | Score (1-5) | Notes |
|-----------|-------------|-------|
| **Frequency**: How often will similar tasks occur? | | |
| **Consistency**: Is the workflow stable or highly variable? | | |
| **Complexity**: Would Claude benefit from structured guidance? | | |
| **Codifiability**: Can steps be clearly documented? | | |
| **Tool-ability**: Are there scripts worth bundling? | | |
| **TOTAL** | /25 | |
**Extraction Recommendation:**
- [ ] **High Priority** (Score 20+): Create skill soon
- [ ] **Medium Priority** (Score 12-19): Accumulate more examples first
- [ ] **Low Priority** (Score <12): Keep as reference only
---
### Proposed Skill Structure (if extracting)
[proposed-skill-name]/ +-- SKILL.md | +-- [Workflow from Section C] +-- scripts/ | +-- [Code from Section E marked as reusable] +-- references/ | +-- schema.md <- [DB knowledge from Section D] | +-- api.md <- [API knowledge from Section D] | +-- examples.md <- [This task as an example] +-- assets/ +-- [Templates from Section F]
### Draft Skill Description
```yaml
name: [proposed-name]
description: |
[Combine trigger profile from Section A with workflow summary from Section C.
This should clearly state WHAT the skill does and WHEN it should activate.]
Knowledge Gaps Identified
| Gap | Impact | How to Fill |
|---|---|---|
| [What you didn't know that would have helped] | [Time lost, errors caused] | [Research, docs, ask someone] |
Patterns Emerging
[If you've done similar tasks before, what patterns are you seeing? This helps identify when multiple SR-PTDs should merge into one Skill.]
### Section J - Tags and Storage
```markdown
## Section J - Tags and Storage
### Tags (for aggregation and skill clustering)
- **Languages**: [Python, TypeScript, SQL, etc.]
- **Frameworks/Libs**: [FastAPI, React, pandas, etc.]
- **Data Stores**: [PostgreSQL, Redis, S3, etc.]
- **External Services**: [Stripe, OpenAI, AWS, etc.]
- **Domain**: [meetings, reports, auth, payments, etc.]
- **Task Pattern**: [CRUD, ETL, generation, transformation, etc.]
- **Security/Compliance**: [PII, encryption, audit, etc.]
### Related Documents
- **Previous similar tasks**: [Links to other SR-PTDs]
- **Existing skills this could extend**: [Skill names]
- **Documentation updated**: [Links]
### Storage
- **Filename**: `SR-PTD_[YYYY-MM-DD]_[task-id]_[brief-description].md`
- **Location**: C:\projects\Skills\Dev_doc_for_skills\
- **Skill Candidate Queue**: [ ] Added to skill backlog
Quick Capture Template (Minimal Version)
For simpler tasks, use this abbreviated format:
# SR-PTD - [Brief Description]
**Date**: YYYY-MM-DD | **Type**: | **Domain**: | **Complexity**:
## Trigger
> [What request/problem initiated this?]
## Workflow (numbered steps)
1.
2.
3.
## Key Decisions
- [Decision] -> [Choice] -> [Why]
## Knowledge Used
- **DB**:
- **API**:
- **Code patterns**:
## Code Written (if reusable)
```python
Output Format (if templatable)
Issues -> Fixes
- [Issue] -> [Fix]
Skill Potential: Low | Medium | High
Notes:
Tags
Languages: | Domain: | Services:
---
## Aggregation: From SR-PTDs to Skills
After accumulating 3-5 SR-PTDs for similar tasks:
### 1. Cluster Related Documents
Group SR-PTDs by:
- Same domain/module
- Similar trigger patterns
- Overlapping workflows
- Shared code/scripts
### 2. Extract Common Patterns
- **Workflow**: What steps appear in ALL instances?
- **Variations**: What steps differ and why?
- **Scripts**: What code was reused or rewritten each time?
- **References**: What knowledge was needed every time?
### 3. Build the Skill
1. Populate SKILL.md from common workflow
2. Add scripts from SR-PTDs marked "reusable"
3. Create reference files from aggregated knowledge
4. Add templates from recurring output formats
---
## Example: Mapping SR-PTD to Skill Components
**Task**: Generated executive summary from meeting transcript
| SR-PTD Section | Content | Skill Component |
|----------------|---------|-----------------|
| Trigger Profile | "Create summary from transcript, Hebrew, action items" | YAML description: "Generate executive summaries from meeting transcripts with action items extraction. Supports Hebrew and English output." |
| Workflow Step 1 | "Parsed transcript with regex for speakers" | SKILL.md: "Step 1: Parse transcript to identify speakers" |
| Workflow Step 2 | "Called Claude API with specific prompt" | scripts/generate_summary.py |
| DB Knowledge | "meetings table schema with transcript_id FK" | references/schema.md |
| API Knowledge | "POST /summaries endpoint format" | references/api.md |
| Output Format | "Executive summary with sections: Overview, Key Points, Action Items" | assets/summary_template.md |
| Issue Fixed | "Hebrew RTL formatting broke in HTML output" | SKILL.md: "Note: Use dir='rtl' for Hebrew output" |
---
## Skill Update Protocol
When a skill already exists and you complete a task using it, evaluate if updates are needed.
**Core Principle:** Every update must pass validation. Default to NOT updating unless criteria are met.
---
### Update Types
**Type A: Fix (Something Broke)**
| Severity | Trigger | Action |
|----------|---------|--------|
| **Critical** | Script crashed/hung, manual intervention required, documented method failed | Must fix immediately |
| **High** | Undocumented edge case, non-obvious workaround discovered | Update recommended |
| **Medium** | Confusing output, ambiguous docs | Consider update |
**Type B: Expand (Capability Growth)**
| Trigger | Action |
|---------|--------|
| New workflow variation discovered | Add conditional path to SKILL.md |
| Reusable code written during task | Extract to scripts/ |
| New schema/API knowledge acquired | Add to references/ |
| Output template created | Add to assets/ |
| Trigger scope should widen | Update YAML description |
**Type C: Skip (No Update Needed)**
- Skill worked as documented
- Environment-specific issue (one-off config)
- User preference (style, not function)
- Information already exists in skill
- Temporary issue (network, load)
- Rare edge case not worth documenting
---
### Decision Tree
Task completed using skill | +-> Did the skill's approach work? | +-> YES -> Any new capabilities discovered? | | | +-> YES -> Type B: Expand | | +-> Run Expansion Checklist | | | +-> NO -> Don't update | +-> NO -> Was manual intervention needed? | +-> YES -> Type A: Fix (Critical) | +-> Must update | +-> NO -> Undocumented edge case? | +-> YES -> Type A: Fix (High) | +-> Run Fix Checklist | +-> NO -> Just confusion? | +-> YES -> Type A: Fix (Medium) | +-> Run Fix Checklist | +-> NO -> Don't update
---
### Pre-Update Validation
**All boxes must be checked. If any fails, STOP.**
#### For Type A (Fix)
- [ ] Gap is **absent**, not just "could be clearer"
- [ ] Verified info doesn't exist in SKILL.md, references/, or examples
- [ ] Issue is **reproducible**, not one-time environmental
- [ ] Solution is **generalizable** to other users/scenarios
- [ ] Benefit **outweighs** added complexity
#### For Type B (Expand)
- [ ] New capability is **reusable** (not one-time)
- [ ] Pattern appeared in **2+ similar tasks** (or clearly will)
- [ ] Doesn't **over-specialize** skill for rare cases
- [ ] Fits within skill's **domain scope**
- [ ] SKILL.md stays **under 500 lines** after addition
---
### Update Mapping: What Goes Where
| What You Learned | Skill Component | Update Action |
|------------------|-----------------|---------------|
| New trigger scenarios | YAML description | Expand trigger phrases |
| Workflow step missing/wrong | SKILL.md body | Fix/add step |
| New workflow branch | SKILL.md body | Add conditional path |
| Error pattern + solution | SKILL.md "Common Issues" | Add issue entry |
| DB schema knowledge | references/schema.md | Add/update tables |
| API endpoint details | references/api.md | Add/update endpoints |
| Code conventions | references/conventions.md | Document patterns |
| Detailed troubleshooting | references/troubleshooting.md | Add scenario |
| Reusable script | scripts/[name].py | Extract and test |
| Script bug fix | scripts/[name].py | Fix and test |
| Output format/template | assets/[template] | Add template file |
### Progressive Disclosure Rules
**SKILL.md body** (loaded when skill triggers):
- Core workflow steps
- Common issues (brief)
- References to detail files
**references/** (loaded on demand):
- Detailed schemas, APIs
- Extended troubleshooting
- Technical deep-dives
- Comprehensive examples
**Rule:** If adding >50 lines to SKILL.md, create a reference file instead.
---
### Update Process
#### Step 1: Classify
- Type A (Fix) or Type B (Expand)?
- Severity level?
- If Type C (Skip) -> Stop
#### Step 2: Validate
- Run appropriate checklist
- All boxes checked? -> Continue
- Any unchecked? -> Stop
#### Step 3: Map
- Identify which skill components need updates
- Use mapping table above
#### Step 4: Update
For each component:
**YAML Frontmatter:**
```yaml
# If trigger scope changed:
description: [Expanded description covering new scenarios]
SKILL.md Body:
# Add to Common Issues (Type A Fix):
### Issue N: [Problem Name]
**Symptom:** [What user sees]
**Cause:** [Why it happens]
**Solution:** [How to fix]
# Add workflow branch (Type B Expand):
## Step N: [Step Name]
- IF [condition] -> [Action A]
- ELSE -> [Action B]
references/[file].md:
# Add detailed knowledge
## [Topic]
[Detailed content that would bloat SKILL.md]
scripts/[name].py:
#!/usr/bin/env python3
"""
Script: [name].py
Purpose: [One line]
Usage: python scripts/[name].py [args]
Returns: 0=success, 1=failure
"""
[Tested, working code]
assets/[template]:
[Reusable template or file]
Step 5: Verify
- YAML frontmatter valid
- SKILL.md < 500 lines
- New scripts tested
- No broken references
- No duplicate information across files
Step 6: Document
Brief summary (2-4 sentences):
- What was updated
- Why (trigger event)
- Which files changed
- Expected improvement
Integration with SR-PTD
When completing a task with an SR-PTD document, extract updates:
| SR-PTD Section | Check For | Skill Update |
|---|---|---|
| Trigger Profile | New trigger phrases | -> YAML description |
| Workflow Executed | Missing/wrong steps | -> SKILL.md body |
| Decision Points | Undocumented branches | -> SKILL.md conditionals |
| Knowledge Accessed (DB) | Schema not in skill | -> references/schema.md |
| Knowledge Accessed (API) | Endpoints not in skill | -> references/api.md |
| Code Written | Marked "reusable" | -> scripts/ |
| Outputs Produced | Marked "template potential" | -> assets/ |
| Issues & Fixes | Novel error patterns | -> SKILL.md Common Issues |
| Skill Extraction Score | 20+ points | -> Consider new skill |
SR-PTD Reusability Flags -> Updates
If SR-PTD contains:
[x] Should become skill script-> Extract to scripts/[x] Could be template-> Extract to assets/[x] Include in Skill (error)-> Add to Common Issues
Skill Update Quick Reference Card
+----------------------------------------------------------+
| SKILL UPDATE QUICK CHECK |
+----------------------------------------------------------+
| |
| UPDATE WHEN: |
| * Script failed/hung (Critical) |
| * Manual intervention required (Critical) |
| * Undocumented edge case hit (High) |
| * Docs caused confusion (Medium) |
| * Reusable code created (Expand) |
| * New knowledge acquired (Expand) |
| |
| DON'T UPDATE WHEN: |
| * Skill worked as documented |
| * Environment-specific issue |
| * User preference (not function) |
| * Info already in skill |
| * Rare edge case |
| |
+----------------------------------------------------------+
| |
| WHERE TO PUT UPDATES: |
| |
| Trigger change -> YAML description |
| Workflow fix/add -> SKILL.md body |
| Error pattern -> SKILL.md Common Issues |
| Detailed knowledge -> references/ |
| Reusable code -> scripts/ |
| Templates -> assets/ |
| |
| RULE: >50 lines -> reference file, not SKILL.md |
| |
+----------------------------------------------------------+
| |
| VALIDATION (all must pass): |
| [ ] Gap is absent, not just unclear |
| [ ] Verified not already documented |
| [ ] Reproducible, not one-time |
| [ ] Generalizable to others |
| [ ] Benefit > complexity |
| [ ] SKILL.md stays < 500 lines |
| |
+----------------------------------------------------------+
Update Examples
Example 1: Type A Fix (Critical)
Situation: Script hung with garbled output, had to kill and use manual commands Update:
- Fix script PowerShell parsing (scripts/kill-port.sh)
- Add Issue 6 to SKILL.md with manual fallback
- Add error pattern to references/troubleshooting.md
Example 2: Type A Fix (High)
Situation: Port showed LISTEN after killing process (zombie connection) Update:
- Add Issue 5 to SKILL.md explaining zombie connections
- Add process-name-based cleanup to workflow
Example 3: Type B Expand
Situation: Wrote reusable validation script during task, SR-PTD marked it "definitely reusable" Update:
- Extract script to scripts/validate_input.py
- Add usage reference to SKILL.md
- Test script independently
Example 4: Don't Update
Situation: Skill worked perfectly, user asked to change output colors Why: Preference, not function. Skill worked as documented.
Quick Reference Card
+----------------------------------------------------------+
| SR-PTD QUICK REFERENCE |
+----------------------------------------------------------+
| |
| AFTER EVERY TASK: |
| 1. Create SR-PTD file in Dev_doc_for_skills/ |
| 2. Fill sections (full or quick capture) |
| 3. Score skill potential |
| 4. Tag for aggregation |
| |
+----------------------------------------------------------+
| |
| SECTION -> SKILL MAPPING: |
| |
| Trigger Profile -> YAML description |
| Workflow Executed -> SKILL.md body |
| Knowledge Accessed -> references/ |
| Code Written -> scripts/ |
| Outputs Produced -> assets/ |
| |
+----------------------------------------------------------+
| |
| SKILL POTENTIAL SCORING: |
| |
| 20+ points -> High Priority (create skill soon) |
| 12-19 points -> Medium (accumulate more examples) |
| <12 points -> Low (keep as reference) |
| |
+----------------------------------------------------------+
| |
| FILE NAMING: |
| SR-PTD_YYYY-MM-DD_[task-id]_[description].md |
| |
| SAVE TO: |
| C:\projects\Skills\Dev_doc_for_skills\ |
| |
+----------------------------------------------------------+
Maintenance
This skill improves based on usage. Updates occur when:
- Documentation structure proves insufficient
- New patterns emerge from aggregated SR-PTDs
- Skill extraction workflow is refined
Updates follow the Skill Update Protocol and require validation.
Score
Total Score
Based on repository quality metrics
SKILL.mdファイルが含まれている
ライセンスが設定されている
100文字以上の説明がある
GitHub Stars 100以上
3ヶ月以内に更新がある
10回以上フォークされている
オープンIssueが50未満
プログラミング言語が設定されている
1つ以上のタグが設定されている
Reviews
Reviews coming soon