Cognitive Gate should be easy to try from three public paths: GitHub, ClawHub, and hosted demos.
Primary install command:
python3 -m pip install "git+https://github.com/somo-ui/cognitive-gate.git@v0.1.6"Smoke test:
cognitive-gate --input "Tidy this room, but don't use the red approach"Expected result: final_action should be blocked because the deterministic mock output violates the red constraint.
Use docs/assets/social-preview.png as the repository social preview image in GitHub settings.
Recommended GitHub topics:
ai-agent
ai-agent-guardrails
ai-safety
llm
llm-guardrails
audit-log
ai-constraints
local-first
model-agnostic
bilingual
Use skill/cognitive-gate/ as the public skill package for ClawHub. A good listing should include:
- Name: Cognitive Gate
- Summary: auditable AI constraints and local decision records
- Tags:
ai-agent-guardrails,ai-constraints,auditable-ai,model-agnostic,local-first - Install: GitHub pip command above
- Skill path:
skill/cognitive-gate/ - Boundary: best-effort reference implementation, not a production security boundary
Do not introduce project-private roadmap names in public directory listings.
ClawHub is the active skill-directory channel. SkillHub upload is intentionally cancelled for the current distribution phase.
ClawHub publish command after logging in:
clawhub skill publish skill/cognitive-gate \
--slug cognitive-gate \
--name "Cognitive Gate" \
--version 0.1.6 \
--tags ai-agent-guardrails,ai-constraints,auditable-ai,model-agnostic,local-first \
--changelog "Initial public skill listing for Cognitive Gate v0.1.6."Use spaces/ as the starter folder for a lightweight Hugging Face Spaces demo. The hosted demo should expose one input box and return:
- final action
- extracted constraints
- route
- blocked reason
- raw local audit record
The demo should not ask for provider API keys. Keep it deterministic so users can reproduce the README examples.