liminal-hq/skills

Shared, multi-agent skills for Liminal HQ workflows.

Goals

• Keep skill logic agent-agnostic and reusable.
• Keep agent-specific instructions thin and explicit.
• Ship versioned, testable skills with clear guardrails.

Repository Layout

• skills/<name>/core/: canonical intent, workflow, and guardrails.
• skills/<name>/adapters/<agent>/: adapter guidance for each agent.
• skills/<name>/references/: templates and supporting documents.
• skills/<name>/tests/: scenario and regression tests for skill behaviour.
• schemas/: metadata schema definitions.
• scripts/: validation and maintenance scripts.

Install (Codex)

Full Skill (recommended)

python3 ~/.codex/skills/.system/skill-installer/scripts/install-skill-from-github.py \
  --repo liminal-hq/skills \
  --path skills/pr-integration \
  --name pr-integration \
  --method git

Conventions

• core/ is the source of truth.
• Adapters must preserve core behaviour and safety gates.
• Skill metadata lives in skill.yaml and must validate against schemas/skill.schema.json.
• skills/index.yaml is the discovery index for installable skills and must stay in sync.
• Only skill-root directories should contain SKILL.md entrypoints.