Tools provided
Tools the agent exposes — your AI client calls them automatically when it needs them.
- open_form — Use this the moment the user wants a health check of a GitHub repository. It returns the questions to ask (the repository, optional comparison repositories, depth) so a form can be shown. Do NOT run any check before calling this; the form is the entry point.
- run_form — Internal: invoked when the user submits the form opened by open_form. Not for direct use.
- run_full — Run a repository health check end to end from live GitHub data: a 0-100 score in a named band with the reason, nine weighted categories each with their formula, strengths and risks citing exact metrics, trend and anomaly flags, one fix-this-first action, a badge snippet and a paste-ready summary. Pass compare_with to compare up to three repositories instead.
- check_repo — Score the health of ONE public GitHub repository (0-100, grade and band with the reason) from live data using a transparent model: activity and 12-week commit trend, issue and pull request response times (median and p90), stale backlog by age, release cadence and regularity, contributor concentration (top-1, top-3, bus factor, Gini) and new-versus-returning contributors, merge versus close ratio, security policy and dependency automation, documentation and license risk. Every finding cites the number and API field it came from; anything GitHub would not return is reported as not checked with a reason, never scored as zero. Use for 'is this repo maintained?' or 'should I depend on this?'.
- compare_repos — Compare the health of 2 or 3 public GitHub repositories side by side: overall score and band per repository, per-category deltas with a tie margin, a like-for-like score over the categories every repository had, trend and anomaly flags, a plain-language verdict and the caveats about sample sizes. Use for 'which of these libraries is better maintained?'.
- explain_score — Walk any part of a repository's health report back to the exact API fields, formulas and weights that produced it: the overall score and band, a category, a single signal, or a flag. Use after check_repo when the user asks 'why is this a 62?' or 'where does that number come from?'. Reuses the report from the last check of that repository when one is still fresh, so it usually costs no GitHub requests.
- list_capabilities — List what this agent checks, how the categories are weighted, its request budgets per depth and its limits. Useful for discovery before running a check.
- plan_inputs — Plan the inputs for a tool: returns the questions, schema and a ready-to-edit example.
- discover_intent — Understand your goal and co-design the exact input via clarifying questions before running.
