Cliare

CLIARE, short for CLI Agent Readiness Evaluation, is an open source Rust tool that audits command line interfaces to see how usable they are for AI agents. It treats a released CLI binary as a black box, probes it under bounded controls, records runtime evidence and infers the real command surface, then emits a machine readable command shape that agent harnesses can load before they start guessing. Its authors describe it as something like OpenAPI for CLIs, generated from observed behavior rather than hand written docs. It serves two audiences. Maintainers get an implementation queue showing where agents lack evidence: missing help, unclear grammar, weak diagnostics, precondition blockers, unvalidated output modes and discovery commands that quietly write files. Agent harnesses and security reviewers can consume the generated command index, shape file and guidance to route commands with known confidence, preconditions and output contracts. The probe run covers help, version, invalid input, output modes and discovered commands within a traversal budget, capturing exit status, bounded output, timing, side effect deltas and runtime context. Outputs include scorecards, issue ledgers, one command summaries, persona reports, CI artifacts and agent skills. CLIARE is free under the Apache 2.0 licence and installs through a release binary installer or Cargo. No paid edition or hosted service is published. The project is maintained by modiqo on GitHub, and its checks focus on four areas: discovery through help text, invocation through argument grammar, consumption through parseable JSON output, and safety through probes that avoid hidden writes.