finds.dev← search

// the find

HKUDS/CLI-Anything

★ 38,855 · Python · Apache-2.0 · updated May 2026

"CLI-Anything: Making ALL Software Agent-Native" -- CLI-Hub: https://clianything.cc/

CLI-Anything is a plugin for AI coding agents (Claude Code, Pi, OpenCode, etc.) that auto-generates structured Python CLI wrappers for arbitrary software by analyzing source code through a 7-phase pipeline. The repo also ships a growing collection of pre-built harnesses for tools like Blender, GIMP, LibreOffice, Audacity, and ~30 others. It's aimed at developers building agentic workflows who need deterministic, scriptable access to software that doesn't have a proper API.

- The 7-phase generation pipeline (analyze → design → implement → test plan → write tests → document → publish) is well-specified in HARNESS.md and produces installable pip packages with a consistent CLI namespace (cli-anything-*), making agent discovery straightforward via standard PATH lookups.

- Pre-built harnesses include real integration with application backends rather than stub implementations — Blender runs actual bpy scripts, Audacity routes through sox, LibreOffice uses headless conversion — so agents get actual software capabilities, not toy reimplementations.

- The SKILL.md per-harness convention gives agents structured discovery metadata (YAML frontmatter, command groups, examples) without requiring any runtime introspection, and the npx skills tooling wires it into a standard discovery flow.

- Security hygiene is visible in the commit history: defusedxml was added for XML parsing, GIMP Script-Fu path injection was fixed, Zoom token permissions were hardened, and the DomShell browser harness has URL validation — this isn't purely vibe-coded glue.

- The generated CLI quality is entirely dependent on what the LLM produces from the source scan — there's no baseline guarantee that a generated harness actually exercises the real application vs. reimplementing thin logic in Python. The 2,269 passing tests count is suspect because many unit tests use synthetic/mocked data rather than a running application.

- Each harness is an independent pip package under cli_anything.*, meaning managing 30+ installed harnesses across projects is a dependency hygiene nightmare with no lockfile or environment isolation story beyond 'pip install -e .'

- The tooling is tightly coupled to Claude Code's plugin marketplace model, and integrations for other agents (Codex, Goose, GitHub Copilot CLI) are marked experimental or community-maintained with no parity guarantees — you're essentially betting on Claude Code staying dominant.

- The monorepo layout (each software gets its own top-level directory with a full agent-harness subtree) doesn't scale well for contribution quality control — harnesses are contributed by community members with varying standards, and the review process is unclear beyond a PR template.

View on GitHub → Homepage ↗

// want more like this?

We dig through GitHub every week and send a few repos picked for what you actually care about — each with an honest take like this one.

Get finds in your inbox → Search again →