Command Claude Code
Hands the P1–P5 findings from the most recent /code-review to Hermit so Claude doesn't burn tokens re-reading every file and retyping every edit. Claude stays on judgment; Hermit does the mechanical apply.
Command Claude Code
Unlike /code-polish, this variant is a distributed state machine. Hermit runs its own review-and-apply loop until it is self-clean, and Claude provides two independent review passes on top as a quality gate. Claude's token spend is capped at 2 review rounds.
Command Claude Code
Commit-and-push are the textbook grunt-work target: short, mechanical, deterministic, and unrelated to reasoning. This skill keeps the deciding parts in Claude and delegates the git commands to Hermit.
Command Claude Code
Same outline as /feature-develop, but the Phase 3 implementation is delegated to Hermit over MCP so Claude's context never absorbs the grunt-work tool output.
Hook Claude Code
Runs before the agent uses a tool for Bash, Edit, Write, NotebookEdit and MultiEdit tool calls, executing block-dangerous-bash.sh and block-during-hermit.sh (2 commands). From cafitac/hermit-agent.
Hook Claude Code
Runs after a tool call finishes for Edit and Write tool calls, executing python-syntax-check.sh. From cafitac/hermit-agent.
Settings file Claude Code
Agent settings declaring 2 hook events (PreToolUse, PostToolUse) and 22 allowed tools.
Instructions file
Instructions for cafitac/hermit-agent, covering hermit project — mcp executor, key modules, public cli, tests and product boundary.
Skill Claude CodeCodex
A read-only audit checklist for Hermit and Claude Code setups, covering how they are structured, used, checked, and improved, including token use and model selection.
MCP server Claude CodeCodexCursor +2
MCP server "codex-channels-local" as configured in cafitac/codex-channels. Runs ./packages/cli/dist/index.js with node. Needs 2 environment variables to run.
Skill Claude CodeCodex
Use this skill when you want to create a real interaction. Run the command first, handle approval/escalation if port binding is required, then guide the user toward pending and reply-latest.
Skill Claude CodeCodex
Use this skill when you want a structured health check for the local runtime. Run the command first, then summarize reachability, interaction count, and the next suggested commands.
Skill Claude CodeCodex
Use this skill when you want low-noise monitoring that can also close the next actionable request once it appears. Run the command first; if reply text is supplied, the command can resolve the next actionable interaction and exit.
Skill Claude CodeCodex
Use this skill when you want the actionable queue only. Run the command first, then summarize the latest pending work.
Skill Claude CodeCodex
Use this skill when the user wants to answer the newest request. If reply text is provided, run the command first. If not, ask for the text or explain the exact command to run.
Skill Claude CodeCodex
Use this skill when you want low-noise monitoring instead of repeatedly polling by hand. Run the command first and only summarize changes that actually happened. This should report the initial state once, then only emit another summary when runtime reachability, actionable count, latest interaction, or next-step…
Skill Claude CodeCodex
Skill "codex-channels" from cafitac/codex-channels, covering codex-channels, execution-first rule, fastest shortcuts, codex operator mode and fastest operator check.
Skill Claude CodeCodex
Use this skill when you want Codex to keep the local loop moving. Run the command first. If it needs reply text, ask for the missing text or tell the user the exact next-step command to rerun.
Skill Claude CodeCodex
Use this skill when you want the fastest high-signal status check before acting. Run the command first, then summarize.