CLIARE CLI
Black-box auditor runs a released CLI, catches side writes to disk and gives agents a command index with a readiness score.
- 441 ★ on the repository
- Last commit
- Apache-2.0
Extended access
Why this level
- Runs the audited tool's binary and, in `host` mode, uses real auth and settings
- Writes measurement artifacts to a chosen folder and changes agent directories when installing skills
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add cli/cliareDetects 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.
- Runs the audited tool's binary and, in `host` mode, uses real auth and settings
- Writes measurement artifacts to a chosen folder and changes agent directories when installing skills
This is third-party code. Review the repository files before installing.
What it does
CLIARE (CLI Agent Readiness Evaluation) runs a built binary of any CLI as a black box and maps the commands from observed behavior: flags, positional arguments, output formats, auth and project-directory requirements. The output is `command-index.json`, a scored summary, an issue ledger and persona reports. Around every probe the tool snapshots the filesystem and flags "safe" commands such as `--help` that quietly write cache or config. The default mode is isolated: a clean environment, temporary directories, limits on time and output size. A separate `cliare skills install` command adds an artifact-review skill to Claude Code, Codex and Cursor.
Who it is for. CLI maintainers preparing a release for agent use, and harness authors who need a verified command map instead of guessing from `--help`.
Good fit when
- A team ships a CLI and wants to see where an agent will stumble: help, diagnostics, parseable output.
- A harness author builds a command index for a third-party tool before wiring it to an agent.
- A security team looks for commands that write files or touch credentials while merely printing help.
- CI needs a gate that blocks a release when the readiness score drops.
Not a fit when
- The program's own logic is under test: CLIARE measures interface shape, not result correctness.
- The target tool cannot be executed even in a sandbox: this tool runs its binary.
- Human-facing documentation is the goal: the reports target agents and maintainers.
Example request
Measure `mycli` with the standard profile and show the summary plus the three issues that hurt an agent most.Limitations
A built binary of the target CLI is required: CLIARE does not analyze source code. The `--execution-mode host` mode takes real auth and configuration from the system, so enable it deliberately and preferably on a test account. Full probes often need fixtures, tokens or a project directory, otherwise some commands end up as `blocked` or `needs_fixture`. The installed skills teach an agent to read an existing artifact folder, and a person or CI starts the measurement. The score describes a release at run time and goes stale after an update, so measurement has to be repeated.
How to disable. Remove the skills from `~/.claude/skills` or `~/.codex/skills`, or the rule from `~/.cursor/rules`, then the binary with `cargo uninstall cliare`. Delete the `.cliare/` result folders by hand.
FAQ
How is measurement different from reading `--help`?
The tool runs the binary and records what happened: real commands, output format, side writes to disk. Help text and docs lag behind a release, while measurement shows current behavior.
Which artifacts appear after a run?
The output folder holds `command-index.json`, `AGENT_SKILL.md`, `persona-harness.json`, the scorecard and the issue ledger. If the artifact map is missing, `cliare describe <folder> --write` builds it.
How do I install the skills into an agent?
`cliare skills list` shows available targets and `cliare skills install --agent all` installs user-level skills for Claude Code, Codex and Cursor. `--dry-run` prints planned writes without changes, and `--scope project` writes into the project directory.
Is it safe to run on a work machine?
The default mode is isolated: a clean environment, temporary directories, time and output limits. `host` mode uses real settings and auth, so it is enabled only on purpose.
Are there install paths besides cargo?
A prebuilt binary comes from the `install.sh` in GitHub Releases with a `SHA256SUMS` check, by default into `$HOME/.local/bin`. There is also `nix profile install github:modiqo/cliare`.
Discussion
No questions or reviews yet.
Related
CLI toolsOver twenty skills for coding agents: MCP and hooks, plan mode, a bug-fix protocol, prompting and a bash safety hook
A Rust CLI for language models from a terminal: an answer, an image, a video, OCR, commit messages and shell commands
A set of 116 skills, 44 agents and safety hooks for Claude Code, Cursor, Codex, Gemini CLI and nine more editors
A CLI that generates and maintains AI context files, CLAUDE.md, Cursor rules, AGENTS.md and Copilot instructions, keeping them accurate as code changes
Apify CLI is the Apify platform console that creates, runs and deploys Actors and prints a ready SKILL.md for coding agents.
A CLI that analyses a repository in six model phases and writes AGENTS.md or CLAUDE.md, with ExecPlan planning on top.
README in short
Two roles: a maintainer gets a fix queue, an agent harness gets a command index. The skills teach an agent to read artifacts, while a person or CI starts the measurement itself.