Clear AI Framework — AI skill for Claude Code

CLEAR - Context, Limits, Evidence, Action, Review

**A vendor-neutral framework for reliable human-AI collaboration.**

CLEAR defines how AI systems should handle context, capabilities, uncertainty, evidence, actions, review, and persistent state across ChatGPT, Claude, Copilot, local models, and agentic systems.

Start here

• CLEAR-v1.0.md - the authoritative v1.0 framework
• QUICK-REFERENCE.md - one-page operational reference
• INVOKE.md - standard invocation and acknowledgement handshake
• CONFORMANCE-CHECKLIST.md - concise checklist for CLEAR-informed/aligned/conformant implementations
• CHANGELOG.md - version history
• RELEASE-NOTES-v1.0.md - v1.0 release notes
• ACKNOWLEDGEMENTS.md - framework lineage and attribution
• LICENSE-DOCUMENTATION - CC BY 4.0 for framework documentation
• LICENSE-CODE - MIT for code, schemas, and reference implementations

The five disciplines

**Context** - use the right information and respect authority. **Limits** - know capabilities, permissions, and uncertainty before acting. **Evidence** - separate facts, derived conclusions, inference, and unknowns. **Action** - distinguish READ, PROPOSE, and EXECUTE. **Review** - verify outcomes, expose limitations, and improve deliberately.

Use CLEAR with an AI

A short invocation is:

Use CLEAR v1.0 for this task. Acknowledge it before beginning and state any material capability limitations.

Expected acknowledgement:

**CLEAR v1.0 acknowledged.** I will apply Context, Limits, Evidence, Action and Review, subject to my governing platform rules and available capabilities.

See **INVOKE.md** for the full portable prompt.

Conformance

Use **CONFORMANCE-CHECKLIST.md** to assess whether an implementation is CLEAR-informed, CLEAR-aligned, or CLEAR-conformant.

Governing maxim

**Interpret human intent generously. Treat evidence, capability, and action exactingly.**

Licensing

**Documentation:** CC BY 4.0 **Code, schemas a