kairos-create-skill
Run the Create KAIROS skill (with KAIROS.md protocol) protocol via KAIROS. The protocol is bundled in references/KAIROS.md. If it is not yet installed in your KAIROS space, mint it first, then run it.
Single vs multiple protocols (aliases)
-
Single protocol: One protocol in references/KAIROS.md . Set metadata.protocol and metadata.protocol_query in SKILL.md.
-
Multiple protocols (aliases): One protocol per alias in references/KAIROS-{alias}.md (e.g. references/KAIROS-workflow-test.md , references/KAIROS-ai-mcp-integration.md ). Set metadata.protocols in SKILL.md to map alias → file path; optionally metadata.protocol_queries to map alias → search query. The agent picks the protocol from the user’s alias/trigger (e.g. /workflow-test → references/KAIROS-workflow-test.md ).
Versioning
-
Skill version: In frontmatter above (metadata.version ). Bumped when the skill instructions or protocol reference change.
-
Protocol version: In each protocol file frontmatter (version ). Bumped when that protocol’s steps or challenges change.
Workflow
-
Search — Call kairos_search with query create KAIROS skill with KAIROS.md protocol (or the user’s phrasing, e.g. "create KAIROS skill", "create skill with protocol"). Use the space parameter if the user or environment specifies one; otherwise omit (default space).
-
If no strong match (e.g. no choice with score ≥ 0.5) — Pick the refine choice from the response and run that protocol to improve the query, then search again.
-
If no match — Read references/KAIROS.md. If it has YAML frontmatter (lines between --- ), pass only the markdown after the closing --- to kairos_mint so the document starts with the H1. Use the same space as for search. Then search again.
-
Run — Follow the chosen match’s next_action : kairos_begin → kairos_next (loop) → kairos_attest . Echo nonces and proof hashes from the server. Do not respond to the user before kairos_attest when must_obey: true .
Protocol summary
The protocol has seven steps: (1) Verify skill directory structure — directory name matches name , SKILL.md present, optional references/ scripts/ assets/. (2) Consult Agent Skills specification via Context7. (3) Gather requirements and draft SKILL.md — purpose, triggers, location, and whether the skill has one protocol or multiple aliases; frontmatter and body. (4) Create protocol(s) via KAIROS — for each alias (or single default), produce a protocol doc and save to references/KAIROS.md or references/KAIROS-{alias}.md . (5) Mint protocol(s) — for each existing references/KAIROS.md or references/KAIROS-*.md , strip frontmatter and call kairos_mint; report chain head URIs or skip. (6) Verify by execution — run the minted protocol(s) or skill in a subagent or new chat; report pass/fail. (7) Lint skill — run skills-ref validate. Use for personal (~/.cursor/skills/) or project (.cursor/skills/) skills.