AgentRules Architect CLI
A CLI that analyses a repository in six model phases and writes AGENTS.md or CLAUDE.md, with ExecPlan planning on top.
- 124 ★ on the repository
- Last commit
Extended access
Why this level
- Repository file excerpts are sent to the chosen cloud models.
- API keys are stored in the local file `~/.config/agentrules/config.toml`.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add cli/agentrules-architectDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Repository file excerpts are sent to the chosen cloud models.
- API keys are stored in the local file `~/.config/agentrules/config.toml`.
This is third-party code. Review the repository files before installing.
What it does
AgentRules Architect analyses a repository with several specialised models and writes the result into an agent rules file: `AGENTS.md` or another name such as `CLAUDE.md`. A six-phase pipeline runs from an initial overview of stack and dependencies through planning, deep per-agent analysis, synthesis and consolidation to the final text. Besides rules, the CLI writes `SNAPSHOT.md`, `.cursorignore` and per-phase snapshots. The `execplan` commands manage long work plans with milestones and a registry under `.agent/exec_plans/`. Phases are assigned models from Anthropic, OpenAI, Google Gemini, DeepSeek and xAI or local Codex and Claude Code runtimes; a researcher can use Tavily web search.
Who it is for. Developers who add AGENTS.md or CLAUDE.md to a large repository and do not want to write agent rules by hand.
Good fit when
- An unfamiliar repository must be described quickly for Codex or Claude Code as a rules file.
- A `CLAUDE.md` file is generated under the needed name with `--rules-filename`.
- `SNAPSHOT.md` is synced after big changes with `agentrules snapshot sync`.
- A multi-day task is driven by an ExecPlan with milestones and registry checks.
Not a fit when
- The repository is a couple of files: a hand-written `AGENTS.md` is faster.
- You want rules hand-written for a team culture: the draft will need rewriting.
- You have no provider keys and the local Codex or Claude Code runtimes are not set up.
Example request
Analyse this repository and create a CLAUDE.md with rules for the agent.Limitations
Python 3.11.9 or newer and access to at least one provider are required: an API key, the local Codex runtime through ChatGPT sign-in, or Claude Code with its own authorization. Repository file excerpts go to the chosen models, and keys are stored in `~/.config/agentrules/config.toml`. Read and edit the generated rules before committing. Researcher web search turns on only with a Tavily key.
How to disable. Remove the package with `pip uninstall agentrules` and the `~/.config/agentrules` directory. The generated `AGENTS.md`, `SNAPSHOT.md` and `.agent/` live in repositories and are deleted by hand.
FAQ
How do I set the rules filename?
With `--rules-filename`, for example `agentrules analyze /path/to/project --rules-filename CLAUDE.md`. A permanent setting comes from `agentrules configure --outputs`.
Are API keys required?
You need at least one way to reach a model: a provider key, ChatGPT sign-in through Codex or Claude Code authorization. Tavily is only for web search.
What is an ExecPlan?
A long work plan with a goal, milestones `MS001`, `MS002` and checklists under `.agent/exec_plans/`. `agentrules execplan new "Title"` creates one.
How do I preview the project tree before analysis?
With `agentrules tree [path]`: it shows an exclusion-aware tree and can save it.
Which files does analysis create?
`AGENTS.md` or a file with your name, `SNAPSHOT.md` by default, plus `.cursorignore`, `.agent/` templates and phase snapshots when enabled.
Discussion
No questions or reviews yet.
Related
CLI toolsA CLI that keeps agent instructions in one place and distributes them to the right config files for Claude Code, Cursor, Codex, Copilot and two dozen other tool
A Claude Code working configuration: a planner with per-milestone review, sub-agents and skills for analysis, refactoring and docs sync
A Rust coding agent with a native GUI, a terminal mode, an ACP mode for Zed and an MCP server
Over twenty skills for coding agents: MCP and hooks, plan mode, a bug-fix protocol, prompting and a bash safety hook
A 2.4 KB charter for CLAUDE.md and three on-demand skills: the agent stops guessing, verifies results and gates irreversible steps
rpa-gen-rules: project rules for agents
rpa-gen-rules
Skill that generates matching project rules for Cursor, Claude Code and Codex built around layered development and BDD
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Repository stars | Main difference |
|---|---|---|---|
| AgentRules Architect (this card) | no | 124 | |
| Caliber | no | 1,310 | caliber creates and maintains context files for a project; AgentRules Architect runs the repository through six analysis phases and manages ExecPlans |
| Rulesync | no | 1,583 | rulesync keeps rules in one source and spreads them across agents; AgentRules Architect writes the rules themselves from repository analysis |
README in short
A package with the `agentrules` console command and the `python -m agentrules` module. An interactive menu walks through analysis, settings and keys, while subcommands cover snapshots, ExecPlans, scaffolds and the project tree.