5.5 KiB
Content Generation Strategies
This skill provides guidance on how to analyze work sessions and generate high-quality content for documentation templates.
Information Sources
Primary: Claude Code Conversation History
This is your richest source of information. The conversation contains:
- Discussions about goals and intent - Why the work was needed
- Design decisions and trade-offs - How approaches were chosen
- Problem-solving iterations - Challenges faced and solutions developed
- Technical details and explanations - Implementation specifics
- Context and reasoning - The "why" behind the "what"
- Future considerations - Next steps and TODOs mentioned
Secondary: Git Repository Data
Use git to supplement and validate the conversation:
- Today's commits:
git log --since="today" --pretty=format:"%h - %s (%an)" --no-merges - Changed files:
git diff --name-only HEAD@{1.day.ago}..HEAD - Commit messages: Capture what was done
- File patterns: Show scope of changes
If no git data is available, rely entirely on conversation context.
Field-by-Field Generation Strategy
Date Field
Auto-fill with today's date in YYYY-MM-DD format. No user prompt needed.
Title/Name Fields
Strategy: Combine project name with work type Sources: User-provided project name + conversation about what was built Example: "Plugin Architecture Refactoring" from project name "plugin-refactor"
Goal/Objective Fields
Strategy: Extract the "why" from the conversation Look for:
- Problem statements at the beginning of the session
- User requests and requirements
- Pain points mentioned
- Desired outcomes discussed
Example: "Restructure the note-taker plugin to align with marketplace architecture principles and improve modularity"
Approach/How Fields
Strategy: Describe the technical approach and methodology Look for:
- Design decisions made during the conversation
- Technologies and tools used
- Architecture patterns discussed
- Implementation steps taken
- Files and components created
Example: "Broke the monolithic command into agents, commands, and skills pattern. Created documentation-assistant agent for expertise, organization-config skill for dynamic discovery, and simplified command for orchestration."
Results/Outcomes Fields
Strategy: Summarize what was successfully accomplished Look for:
- Features or code completed
- Problems solved
- Artifacts created
- Tests passing
- Successful builds or deployments
Example: "Successfully restructured plugin with 4 new files: agent, 2 skills, updated command. Reduced command from 189 lines to ~60. Implemented dynamic template discovery."
Challenges/Blockers Fields
Strategy: Identify difficulties and how they were resolved Look for:
- Errors encountered
- Unexpected behaviors
- Decisions that required debate
- Problems that needed creative solutions
- Anything marked as "tricky" or "challenging"
Example: "Initial approach used hardcoded template lists. Refactored to use Glob for dynamic discovery to improve maintainability."
Next Steps/Future Work Fields
Strategy: Extract forward-looking items Look for:
- TODOs mentioned
- Features deferred
- Ideas for improvements
- Follow-up tasks identified
- Testing or validation needed
Example: "Update documentation to reflect new structure. Test template discovery with actual files. Consider adding validation for template fields."
Content Generation Process
Step 1: Analyze the Full Session
Review the entire conversation history to understand:
- What was the initial request?
- What was discussed and decided?
- What was implemented?
- What challenges came up?
- What was the outcome?
Step 2: Supplement with Git Data
If available, use git commits to:
- Validate what files were changed
- See commit messages for additional context
- Understand scope of changes
Step 3: Generate Field Content
For each template field:
- Identify what type of information is needed
- Extract relevant details from conversation + git
- Synthesize into clear, specific content
- Make it detailed enough to be useful
Step 4: Present for Approval
Format:
[Field Name]:
Your proposed content here...
Accept this? (Press Enter to accept, or type your changes)
Default behavior is acceptance - make content good enough to just press Enter.
Quality Guidelines
Be Specific:
- Use actual file names, feature names, technical terms
- Include numbers when relevant (lines of code, number of files, etc.)
- Reference specific technologies and approaches
Be Comprehensive:
- Capture both what and why
- Include context that might be forgotten later
- Don't just list actions, explain reasoning
Be Concise:
- Focus on important details
- Avoid unnecessary verbosity
- Make every sentence count
Be Accurate:
- Base proposals on actual conversation and git data
- Don't invent or assume information
- If uncertain, ask the user
Example: Full Template Fill
Template Field: What did you build? Generated Content: "Restructured the note-taker plugin from a single 189-line command into a modular architecture with agent, commands, and skills. Created documentation-assistant agent for note-taking expertise, organization-config skill for dynamic template discovery, and content-generation skill for analysis strategies. Reduced command file to ~60 lines focused on workflow orchestration."
Why this is good:
- Specific numbers (189 lines → 60 lines)
- Named components created
- Explained the transformation
- Clear outcome
- Based on actual conversation