Agentic OS workflow
Agentic OS adds a five-step workflow and checks in git hooks and CI so that a step without evidence does not count as done
- 234 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- The installer copies rule, script and hook directories into the project, and checks run scripts on your machine and read git diffs.
Install
- Works in
Manual install
./agentic-os/installers/deploy_brain.sh /path/to/your-projectFirst `git clone https://github.com/KbWen/agentic-os.git`. Add `--dry-run` for a trial run.
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The installer copies rule, script and hook directories into the project, and checks run scripts on your machine and read git diffs.
This is third-party code. Review the repository files before installing.
What it does
Agentic OS is a governance layer for agents delivered as rules and scripts. It runs work through five steps: plan, build, review, test and ship, and no step counts as done without recorded evidence. The `AGENTS.md` file and directories of rules, skills and workflows are copied into a project by the `installers/deploy_brain.sh` installer, which can first run with `--dry-run`. On top of ordinary rules it adds checks the agent does not control. The `scan_credentials.py` script looks for leaked secrets in files, the git index or a commit range with redacted output. The `validate.sh` script reads the task work log and fails when a phase is skipped or evidence is missing. The `record_test_run.py` script runs the test suite and records the exit code and a code fingerprint. The pre-commit hook and the CI job are switched on by hand, and the installer prints both lines. There are `/bootstrap`, `/review`, `/ship`, `/spec-intake`, `/audit` and `/handoff` commands, 14 expert checklists by task type and a work log for session handoff.
Who it is for. Developers who give agents large tasks and want the word done backed by secret scans, tests and recorded phases.
Good fit when
- The agent declares a task done with no tests and you need a gate before shipping.
- A secret should fail the commit and not reach git history.
- An existing repository adopts agent discipline: the read-only `/audit` is safe to start with.
- Work passes to a new session and a resume block via `/handoff` is needed.
Not a fit when
- You need protection against a deliberately faked log: the check reads the agent record and will not catch a fabricated one.
- One rules file is enough: Agentic OS adds directories, scripts and hooks.
- The project has no git or CI: the main gates live in commit hooks and CI jobs.
Example request
Read `AGENTS.md` and follow it. Do not claim completion until `/review` and `/test` pass. Start with `/bootstrap` for this task.Limitations
Checks work only after you enable them: the pre-commit hook and the CI job must be connected by hand and the installer prints both lines. Work logs sit in git-ignored files, so CI never sees them and `validate.sh` should run on the developer machine or in a commit hook. The validator checks the record the agent keeps and catches a skipped phase but not a faked record or a task with no record. If the agent ignores the workflow or a checklist, no machine check catches it and that part stays advisory like any rules file. Existing `AGENTS.md` and `CLAUDE.md` are kept and the framework version lands beside them as `.acx-incoming`.
How to disable. Remove the `.agent` and `.agentcortex` directories and rule blocks added by the installer, plus the commit hook and CI job if you enabled them.
FAQ
How do I install Agentic OS?
Clone the repository and run `./agentic-os/installers/deploy_brain.sh /path/to/your-project`. You can first run it with `--dry-run` to see what changes.
How do I see the secret scan without installing?
Run `bash demo/run.sh`, on Windows `pwsh demo/run.ps1`. The demo shows `scan_credentials.py` catching a leaked key.
Which agents are supported?
Claude Code, Codex, Cursor, Copilot, Antigravity and any agent that reads Markdown, through `AGENTS.md`.
Where do I start in a new and an existing project?
In a new one start with `/spec-intake`, in an existing one with `/audit`, and for a single task `/bootstrap` fits.
Does install touch my `AGENTS.md` and `CLAUDE.md`?
No. Existing files are kept and the framework version lands beside them as `.acx-incoming` to merge.
Discussion
No questions or reviews yet.
Related
WorkflowsAGENTS.md instructions and skills with a test-first process that scales to the task, for Cursor, Codex CLI, Gemini CLI.
A workflow template for Claude Code and five more agents: a backlog task, a diff review and a commit.
Agentplane handles task state, branches and result evidence in a CLI while the coding agent focuses on the content of the change.
A library of 30-plus instruction skills for coding agents: planning, building, testing, hallucination guards and shipping
Over twenty skills for coding agents: MCP and hooks, plan mode, a bug-fix protocol, prompting and a bash safety hook
A skill that analyzes a repository and generates files for AI agents: AGENTS.md, pointers, reviewers, CI and templates
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Repository stars | Main difference |
|---|---|---|---|
| Agentic OS (this card) | no | 234 | |
| Agent Governance Toolkit | yes | 6,562 | Agent Governance Toolkit installs as a Claude Code plugin, while Agentic OS works through `AGENTS.md`, scripts and git hooks for many agents. |
| step-gate | no | 945 | step-gate holds the order of steps, while Agentic OS adds secret scanning, test run records and a work log. |
README in short
The README shows the difference between a rules file and enforcement, a quick start, lists of commands, checklists and the project own tests. MIT license, with CI and security checks.