The Gibson
The-AIE/the-gibson
Governance harness with a gated plan-to-ship pipeline for agent fleets on Claude Code, Codex, Grok or Hermes
The Gibson is a portable, self-improving SDLC harness that a fleet of coding agents runs under. You describe what you want, and the fleet plans, builds, tests, reviews and ships it through gates you can audit, on any runtime including Hermes Agent.
What The Gibson does
The Gibson is a doctrine-and-tooling repository installed into a target project rather than a hosted service. It is deliberately agent-agnostic: the canonical harness is plain Markdown, shell scripts and GitHub Actions, which any runtime, including Claude Code, OpenAI Codex, Grok, Hermes and pi, can read and any CI can enforce, with vendor-specific ergonomics kept as thin adapters over that same core. Work moves through a pipeline of plan, decompose into issues, build, test, review, a UI/UX evaluation against live deployments, a security scan, merge, deploy and a retrospective, and it does not stop short of a human gate unless one is required.
Three layers make up the system: doctrine in AGENTS.md and role or job playbooks, deterministic enforcement gates in CI and scripts that do not care who wrote the code, and a versioned memory of lessons, decisions and incidents. Every time the system asks a human for something, the README's Ask Contract requires it to state what it is asking, what it does, why, and the risks, in plain language. The Gibson is described as the harness layer, working alongside a separate project called Mission Control that acts as the runtime control plane deciding who works.
Key features
- Plan-to-ship pipeline with human gates: plan, build, test, review, UI/UX eval, security, deploy
- Agent-agnostic doctrine in plain Markdown, shell scripts and GitHub Actions
- Versioned memory of lessons, decisions and incidents under memory/
- Ask Contract requiring plain-language, risk-stated requests to humans
- Playwright-based UI/UX evaluation against live deployments
- Fork-and-upstream workflow that keeps a customized copy updated
When to use it
- Enforcing one shared, auditable SDLC process across a fleet of different coding agents
- Requiring a human security and UX gate before a deployment ships
- Forking the harness for a team's own conventions while still pulling upstream improvements
Who it is for: Teams running multiple coding agents, including Hermes Agent, who want one enforced software delivery process across all of them.
How it fits with Hermes Agent
Names Hermes Agent as one of the runtimes it is agent-agnostic across, reading the same plain-Markdown doctrine as Claude Code, Codex and Grok.
FAQ
What is The Gibson?
The Gibson is a portable SDLC harness and governance doctrine for agent fleets. It takes a plan through build, test, review, security and deployment under human-auditable gates.
Does The Gibson work with Hermes Agent?
Yes. The README lists Hermes Agent as one of the runtimes it supports, since the core harness is plain Markdown, shell scripts and GitHub Actions that any agent can read.
Is The Gibson free and open source?
Yes. The repository is released under the Apache-2.0 license.
Similar dev tools for Hermes Agent
All dev toolsRust CLI that backs up and restores Hermes Agent and OpenClaw installations to tar.zst archives
huangrichao2020 hermescheckArchitecture and runtime health checks for Hermes Agent checkouts, forks and deployments
0xNyk openclaw-to-hermesMigration tool that moves an OpenClaw setup to Hermes Agent, with dry-run and backup
nicolasmelo1 LogionDeveloper tooling for a registry that measures whether AI-agent skills and plugins actually work
BULDEE AI Craftsman SuperpowersArchitecture gate for AI coding agents that blocks writes breaking DDD and Clean Architecture rules
vector4wang EdgeGlowmacOS menu bar app that lights up your screen edges while Claude Code or Hermes Agent is working
Related guides: How to install Hermes Agent · Run multiple Hermes agents with profiles