The fastest way to contribute: add a new thinking framework agent.
Drop a .md file in .claude/agents/ following this template:
---
name: your-framework
description: One-line description of what this framework does
---
You are a **Your Framework** thinking agent.
Your job is to [core purpose — one sentence starting with a verb].
## Inputs
You will receive:
- **CONTEXT**: Current state, what exists, what's been tried
- **SUBJECT**: The problem, approach, or output to analyze
- **INPUT_FILES**: Specific file paths to read for evidence
## Your Task
Apply [Your Framework] by answering these questions:
### Question 1: [Your first question]
[Details about what to assess]
### Question 2: [Your second question]
[Details]
### Question 3: [Your third question]
[Details]
## Output Format
For each finding:
**KEY_TERM** → [description]
**EVIDENCE** → [citation: file_path:location or CONTEXT reference]
**CONCLUSION** → [what this means]
## Verdict
End with exactly one of:
- **PASS** — [condition for passing]
- **FAIL** — [condition for failing]
Include:
- `findings`: List of specific findings
- `recommendations`: Specific actions to take
## Evidence Requirement
Every claim must cite a source. [Describe your specific citation rules. If a claim cannot be cited, state the appropriate flag.]Run the validation script:
bash scripts/validate-structure.shYour new agent should show green checkmarks for frontmatter and evidence requirement.
Your framework is available via:
/consider your-framework "the problem to analyze"
No registration, no config changes, no orchestrator modification needed.
- 3 core questions (can deviate if methodology demands it)
- Binary verdict: PASS or FAIL (nuance goes in findings, not verdict)
- Evidence required: Every claim cites a source — this is non-negotiable
- Domain-agnostic: No references to specific projects, technologies, or domains
- 55-80 lines: Keep it focused. If longer, you're probably combining two frameworks.
- Independent: Your framework must work without knowing other frameworks exist
| If your framework... | Category |
|---|---|
| Generates ideas or explores possibilities | Exploration |
| Breaks down problems to fundamentals | Deconstruction |
| Evaluates or selects approaches | Strategy |
| Checks implementation against reality | Verification |
| Simplifies or removes | Refinement |
| Audits AI-generated work specifically | Agentic Audit |
- Fork the repository
- Create a feature branch (
git checkout -b feature/your-feature) - Make your changes following the conventions above
- Commit with a descriptive message (
git commit -m "Add: description of change") - Push to your fork (
git push origin feature/your-feature) - Open a Pull Request against
main
Pull requests should include:
- Description of what changed and why
- Any testing you performed
- Reference to related issues (if applicable)
Open an issue on GitHub with:
- Framework name
- What happened vs. what you expected
- The command you ran