Skill Claude CodeCodex
Declare and call LLM backends in mhl. Use for agent, router, prompt, skill and schema declarations, before and after hooks, structured output, and retry, cache, ratelimit or fallback configuration.
Skill Claude CodeCodex
Core syntax, types and semantics of mhl (Meta-Harness Language, .mh files), a language most models have not seen. Load before writing, editing, reviewing or debugging any .mh file, or when mhl lint reports a parse, scope or type error.
Skill Claude CodeCodex
Expose mhl workflows to other agents with mhl serve. Use for MCP tools, A2A skills, the mhlrun async tools, input and output contracts for callers, human-in-the-loop resume, --token and --state-dir hardening, and mhl pack.
Skill Claude CodeCodex
Test and validate mhl programs. Use for test and describe blocks, mocking agents without calling an LLM, asserting workflow outcomes with Name.run, and reading mhl lint, mhl test or mhl run --dry-run results.
Skill Claude CodeCodex
Give mhl programs side effects and integrations. Use for tool declarations, native ops (cmd, fs, git, http, json, time), memory stores, extension mcp and extension a2a clients, and credentials through env().
Skill Claude CodeCodex
Design mhl pipeline and workflow declarations. Use for step structure, pipeline vs workflow, goto and route, loop and repeat, parallel, spawn and wait, pause and resume, checkpoints, typed signatures, output projection and lifecycle hooks.
At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: