# Intrikata Topology skills — agent adapter

For agentic harnesses other than Claude Code (Codex/AGENTS.md-family, Cursor, Gemini, ...):

1. Discover: fetch `https://topology.intrikata.com/skills/index.json` — every skill with name, description, and a raw URL.
2. Fetch any skill directly: `https://topology.intrikata.com/skills/<name>.SKILL.md`. Each file is self-contained
   markdown instructions with YAML frontmatter (`name`, `description` = trigger contract).
3. Install wholesale: `git clone https://topology.intrikata.com/git/intrikata-topology-skills.git` and mount
   `intrikata-topology/skills/` wherever your harness loads instruction files; treat each SKILL.md
   as a conditionally-loaded instruction module keyed on its frontmatter description.
4. Claude Code users: see `https://topology.intrikata.com/install.md` for the marketplace path.

Skills are written in Claude Code dialect (tool names like Read/Grep/Bash, `Skill` invocation).
Map primitives accordingly; the frontmatter contract (when to fire) is harness-neutral.
Verify integrity against `MANIFEST.json` (sha256 per file) in the package root.