hathach/agentrc
Agent-neutral personal config for skills, commands, and hooks shared by Claude Code and Codex
ARCHETYPE
Orchestrator
A bench of specialised subagents. The main agent mostly delegates.
Copy this rig
# review before running: this installs third-party code $ npx degit hathach/agentrc/agents ./rig-agentrc/agents $ npx degit hathach/agentrc/skills ./rig-agentrc/skills
MCP servers are added to Claude Code at local scope; env vars are shown as YOUR_… placeholders — we never store values. Files are fetched with degit into a separate folder so you can review before merging.
$ npx degit hathach/agentrc/skills/ci-rerun .claude/skills/ci-rerun $ npx degit hathach/agentrc/skills/cowork .claude/skills/cowork $ npx degit hathach/agentrc/skills/download-doc .claude/skills/download-doc $ npx degit hathach/agentrc/skills/esp-target-debug .claude/skills/esp-target-debug $ npx degit hathach/agentrc/skills/etm-trace .claude/skills/etm-trace $ npx degit hathach/agentrc/skills/followup-issue .claude/skills/followup-issue $ npx degit hathach/agentrc/skills/headless-chief .claude/skills/headless-chief $ npx degit hathach/agentrc/skills/herdr-peer .claude/skills/herdr-peer $ npx degit hathach/agentrc/skills/pr-babysit .claude/skills/pr-babysit $ npx degit hathach/agentrc/skills/pr-reply .claude/skills/pr-reply $ npx degit hathach/agentrc/skills/pr-review .claude/skills/pr-review $ npx degit hathach/agentrc/skills/read-doc .claude/skills/read-doc $ npx degit hathach/agentrc/skills/read-pcb .claude/skills/read-pcb $ npx degit hathach/agentrc/skills/rtt .claude/skills/rtt $ npx degit hathach/agentrc/skills/simplify-gate .claude/skills/simplify-gate $ npx degit hathach/agentrc/skills/target-debug .claude/skills/target-debug $ npx degit hathach/agentrc/skills/usb-kernel-debug .claude/skills/usb-kernel-debug $ npx degit hathach/agentrc/skills/usb-sniffer .claude/skills/usb-sniffer $ npx degit hathach/agentrc/skills/worktree-reset .claude/skills/worktree-reset
$ curl -fsSL --create-dirs -o .claude/agents/chief.md https://raw.githubusercontent.com/hathach/agentrc/main/agents/chief.md $ curl -fsSL --create-dirs -o .claude/agents/code-verifier.md https://raw.githubusercontent.com/hathach/agentrc/main/agents/code-verifier.md $ curl -fsSL --create-dirs -o .claude/agents/code-writer.md https://raw.githubusercontent.com/hathach/agentrc/main/agents/code-writer.md $ curl -fsSL --create-dirs -o .claude/agents/coworker.md https://raw.githubusercontent.com/hathach/agentrc/main/agents/coworker.md $ curl -fsSL --create-dirs -o .claude/agents/finding-verifier.md https://raw.githubusercontent.com/hathach/agentrc/main/agents/finding-verifier.md $ curl -fsSL --create-dirs -o .claude/agents/hil-operator.md https://raw.githubusercontent.com/hathach/agentrc/main/agents/hil-operator.md $ curl -fsSL --create-dirs -o .claude/agents/hw-debugger.md https://raw.githubusercontent.com/hathach/agentrc/main/agents/hw-debugger.md $ curl -fsSL --create-dirs -o .claude/agents/hw-validator.md https://raw.githubusercontent.com/hathach/agentrc/main/agents/hw-validator.md $ curl -fsSL --create-dirs -o .claude/agents/pr-ci-watcher.md https://raw.githubusercontent.com/hathach/agentrc/main/agents/pr-ci-watcher.md $ curl -fsSL --create-dirs -o .claude/agents/pr-review-validator.md https://raw.githubusercontent.com/hathach/agentrc/main/agents/pr-review-validator.md $ curl -fsSL --create-dirs -o .claude/agents/pvs-studio.md https://raw.githubusercontent.com/hathach/agentrc/main/agents/pvs-studio.md $ curl -fsSL --create-dirs -o .claude/agents/relay.md https://raw.githubusercontent.com/hathach/agentrc/main/agents/relay.md
This rig commits no guardrails. Here is the community baseline instead — the deny/ask rules most often found across all 7,204 rigs:
{
"permissions": {
"deny": [
"Read(./.env)",
"Read(**/.env)",
"Read(~/.ssh/**)",
"Bash(rm -rf *)",
"Read(**/*.pem)",
"Bash(rm -rf /)",
"Bash(git push --force:*)",
"Bash(sudo *)",
"Read(.env)",
"Bash(rm -rf /*)",
"Read(./.env.*)",
"Read(~/.aws/**)",
"Bash(git push --force*)",
"Bash(rm -rf:*)",
"Read(**/*.key)",
"Read(**/.env.*)",
"Bash(sudo:*)",
"Bash(git reset --hard*)",
"Bash(git reset --hard:*)",
"Read(.env.*)"
],
"ask": [
"Bash(git push:*)",
"Bash(git push *)",
"Bash(git commit:*)",
"Bash(rm *)",
"Bash(rm:*)",
"Bash(git rebase *)",
"Bash(wget *)",
"Bash(npm publish:*)",
"Bash(git commit *)",
"Bash(gh pr merge *)"
]
}
} Skills (19)
ci-reruncoworkdownload-docesp-target-debugetm-tracefollowup-issueheadless-chiefherdr-peerpr-babysitpr-replypr-reviewread-docread-pcbrttsimplify-gatetarget-debugusb-kernel-debugusb-snifferworktree-reset
Subagents (12)
| chief model: opus | Dispatch-only main session for an existing task worktree. Start it inside the worktree with `claude --agent chief`. |
| code-verifier model: opus | Review one directory or one diff against one review dimension (correctness, concurrency, datasheet/errata conformance, style) with coverage-first structured findings. Read-only; refutation of a single |
| code-writer model: opus | Implement one well-scoped change in one assigned scope (a directory or an explicit file set), following the repository's style, verified by the build command the prompt names. Use for fan-out developm |
| coworker model: sonnet | Run one cowork.py operation against Codex in this checkout and return its output unchanged. |
| finding-verifier model: opus | Adversarially verify one review finding ("try to refute this") or one fix ("does this diff address finding X?") with a yes/no answer in the JSON shape the prompt names. Read-only; default verdict is r |
| hil-operator model: sonnet | Run one hardware-in-the-loop action on a project's physical test rig - lock, flash, test battery, report - exactly as the project's `HIL contract:` defines it. Strictly one instance at a time. Never e |
| hw-debugger model: opus | Diagnose and fix one firmware symptom on real hardware - a hypothesis loop inside one locked round, then, with the board released, finalize and commit the supported fix. Holds the board lock, flashes |
| hw-validator model: opus | Check one claim about firmware behaviour on real hardware with target-side evidence - reproduce a failure, prove a committed fix, place a CI failure - then restore the board and checkout. The only rol |
| pr-ci-watcher model: opus | Judge one PR's failing CI checks from the evidence collect.py gathered, classify each failure (infra flake / real / rig-side / unclassified), re-run infra ones once, report every failure with its firs |
| pr-review-validator model: opus | Harvest one PR's bot reviews (Copilot, CodeRabbit, Greptile, code scanning), report where each auto-running bot stands on the head SHA, and adversarially validate each finding against the code — verdi |
| pvs-studio model: sonnet | Run PVS-Studio static analysis (SAST + MISRA C:2023 / C++:2008) over a compile_commands.json and report structured findings, gated on diagnostics in files changed vs a base ref. Read-only; never edits |
| relay | A workflow's script relay - runs the command its prompt names and returns its output as the schema asks. |
Similar rigs
lee-to/aif-handoff
Autonomous Kanban board where AI agents plan, implement, and review your tasks — fully hands-off.
Pragmatist 4.5k tok ·
workflow/dotfiles
💫 NixOS configuration files
Minimalist 235 tok ·
teabranch/open-responses-server
Wraps any OpenAI API interface as Responses with MCPs support so it supports Codex. Adding any missing stateful features. Ollama and Vllm compliant.
Minimalist 1.3k tok ·
AdityaRon/claude-code-harness
Machine-level security, audit, and context hooks for Claude Code
Minimalist 236 tok ·