// Open-source catalog for AI coding harnesses
Drop-in skills and agent adapters for Claude Code, OpenAI Codex, GitHub Copilot, Cursor, Gemini CLI, Kiro, and Lovable.
/plugin marketplace add TechTideOhio/techtide-harness-kit// the catalog
// what ships
skills/Portable task workflows with frontmatter.
agents/Specialist roles with harness adapters.
rules/Harness-specific operating guidance.
mcp/Trusted MCP server notes.
catalog/Machine-readable indexes + trust metadata.
powers/Kiro Power packages.
plugins/Codex plugin packages + templates.
$ ls techtide-harness-kit/skills agents rules mcp catalog powers plugins
// supported harnesses
| Harness | Command | |
|---|---|---|
| Claude Code | /plugin marketplace add TechTideOhio/techtide-harness-kit | |
| OpenAI Codex | codex plugin marketplace add TechTideOhio/techtide-harness-kit | |
| GitHub Copilot | copilot plugin marketplace add TechTideOhio/techtide-harness-kit | |
| Cursor | git clone … && add as plugin directory | |
| Gemini CLI / Antigravity | npx thk-export-agents --platform gemini --all | |
| Kiro | add selected powers/techtide-* directories | |
| Lovable | npm run lovable:write # one ZIP per skill |
// install
Clone, install dependencies, validate the catalog, then export agents to your harness of choice.
$ git clone https://github.com/TechTideOhio/techtide-harness-kit$ cd techtide-harness-kit && npm install$ npm run validate$ npx thk-export-agents --help
// trust posture
Read-first, approval-gated, target-confirmed for risky ops.
Skills carry tool scopes, data classes, network posture, approval gates, audit metadata.
External skills are quarantine-first.
Compliance mappings are engineering aids, not legal attestations.
See TRUST.md, EVALS.md, DATA-HANDLING.md, PROMPT-INJECTION.md.