Menu
Capture and compound learnings from conversations into reusable commands and skills through interactive dialogue
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Based on SOC occupation classification
Autonomously optimize any Claude Code skill by running it repeatedly, scoring outputs against binary evals, mutating the prompt, and keeping improvements. Based on Karpathy's autoresearch methodology. Use when: optimize this skill, improve this skill, run autoresearch on, make this skill better, self-improve skill, benchmark skill, eval my skill, run evals on. Outputs: an improved SKILL.md, a results log, and a changelog of every mutation tried.
Extract clean markdown content from web pages using Defuddle CLI, removing clutter and navigation to save tokens. Use instead of WebFetch when the user provides a URL to read or analyze, for online documentation, articles, blog posts, or any standard web page.
Create and edit JSON Canvas files (.canvas) with nodes, edges, groups, and connections. Use when working with .canvas files, creating visual canvases, mind maps, flowcharts, or when the user mentions Canvas files in Obsidian.
Create and edit Obsidian Bases (.base files) with views, filters, formulas, and summaries. Use when working with .base files, creating database-like views of notes, or when the user mentions Bases, table views, card views, filters, or formulas in Obsidian.
Interact with Obsidian vaults using the Obsidian CLI to read, create, search, and manage notes, tasks, properties, and more. Also supports plugin and theme development with commands to reload plugins, run JavaScript, capture errors, take screenshots, and inspect the DOM. Use when the user asks to interact with their Obsidian vault, manage notes, search vault content, perform vault operations from the command line, or develop and debug Obsidian plugins and themes.
Create and edit Obsidian Flavored Markdown with wikilinks, embeds, callouts, properties, and other Obsidian-specific syntax. Use when working with .md files in Obsidian, or when the user mentions wikilinks, callouts, frontmatter, tags, embeds, or Obsidian notes.
| name | compound |
| description | Capture and compound learnings from conversations into reusable commands and skills through interactive dialogue |
Capture learnings from conversations and compound them into reusable commands, skills, and documentation through an interactive, one-shot workflow.
Compound engineering means each unit of work makes the next unit easier. This skill helps you:
Ask questions with lettered options for quick answers:
1. What type of improvement is this?
A. New command pattern
B. Update existing command
C. New skill guidance
D. Update AGENTS.md
E. Other: specify
2. Which proposals should be applied?
A. All proposals
B. Specific proposals (e.g., "A and C")
C. None
Each proposal should include:
## Proposal A: Add validation pattern to commands
### Summary
Add validation guidance to command templates.
### Target
- `commands/template.md` (new or existing)
### Patch
```diff
+## Validation
+
+Always validate inputs before processing.
+Use the existing validation patterns in the codebase.
## Target Files
Proposals can update:
- `commands/` - Slash commands in the opencode config
- `skills/` - Skill definitions in the opencode config
- `AGENTS.md` - Global agent guidance
- Any explicitly specified file path
## Extraction Rules
When analyzing context for proposals, look for:
1. **Repeated patterns**: "I keep doing X the same way" → propose a command or skill update
2. **Documentation gaps**: "This was hard to figure out" → propose AGENTS.md update
3. **Workflow inefficiencies**: "This takes too many steps" → propose a new command
4. **Knowledge loss**: "I learned this but will forget" → propose documentation
## When to Generate Proposals
Generate proposals **only** when:
- There is a clearly actionable improvement
- The user explicitly requests compounding
- A pattern has been observed multiple times
- The learning has direct implications for commands/skills behavior
Do **not** generate proposals for:
- Generic reflections without actionable follow-ups
- One-off issues that don't indicate a pattern
- Items that are already well-documented
## Workflow
### Interactive Compound Flow
1. User provides context: `/compound [description]`
2. Command asks clarifying questions if needed
3. Command presents 1-3 proposals
4. User selects proposals to apply (e.g., "A and C")
5. Command applies patches directly to files
6. Command outputs summary of changes
### Example Session
User: /compound I want to add a pattern for error handling
Assistant:
User: A
User: A
Assistant: Proposal A: Add error handling pattern to command template
Summary: Add error handling guidance to command templates.
Target: commands/template.md
Patch:
+## Error Handling
+
+Always handle errors gracefully.
+Provide clear error messages for the user.
+Log errors for debugging.
Applied 1 proposal(s):
commands/template.md: Add error handling guidanceFiles modified:
commands/template.md
## Best Practices
- Capture learnings promptly while context is fresh
- Propose small, focused improvements rather than large rewrites
- Ask clear, actionable questions with lettered options
- Present diff-style patches for transparency
- Apply changes directly without intermediate files
- Output clear summaries of what was changed