toolbox

What problem does it solve? Maintaining a shared toolbox of agent-facing CLI tools can drift into inconsistency: tools with noisy output, missing help text, or unclear purposes waste tokens and confuse agents. This Skill provides the conventions and end-to-end checklist for deciding when a tool deserves to exist and how to add, rename, or remove one cleanly in the workflows repo. ## Core Features & Use Cases - Tool admission criteria: Add a tool only when the task repeats, ad-hoc shell work burns tokens, and the result is deterministic enough to freeze into a command. - Tool contract: Enforces rules such as living in bin/ with no extension, terse line-oriented output, short --help that exits 0, errors on stderr with non-zero exit codes, and working from any directory. - End-to-end workflow: Covers writing the executable, wiring dependencies into requirements.txt, package.json, or install.sh, authoring the tool's own SKILL.md, updating the README table, and verifying with ./install.sh --check. - Use Case: When you notice an agent repeatedly running the same multi-step shell pipeline to inspect git history, use this Skill to decide whether it merits a new bin/ tool and follow the checklist to ship it consistently. ## Quick Start Ask the agent to add a new command-line tool to the workflows repo following the toolbox conventions, including its SKILL.md, README row, and install verification.