Tools provided
Tools the agent exposes — your AI client calls them automatically when it needs them.
- optimize_pipeline — Find where your CI pipeline spends its wall clock and propose the deterministic fixes: broken or missing dependency caches, steps that could run concurrently, oversized test jobs worth sharding, and redundant steps. Returns a before/after duration, a per-step share breakdown and a workflow diff preview. Savings are excluded from the total when the step already runs in parallel, or when the proposal rests on a field you never declared.
- forecast_pipeline_health — Read your CI run history and say where the pipeline is going: the per-stage trend fitted by least squares, the stage that will consume the most of the run first and when it overtakes today's largest, and whether the growth is work time or queue time. Every figure is computed from the runs you supply; a stage with too little history gets no projection and is named with how many runs it is short by.
- plan_deployment — Turn a change, a set of environments and your own deployment record into a staged rollout: the order, the gate on each stage and why that gate, the signals to watch traced back to the risks you declared, the stop conditions, and the soak time. Gates come from your measured failure rate per environment; a stage with too little history or no supplied threshold is named as a gap rather than given a number nobody measured. It plans only — it deploys nothing.
- plan_rollback — Work out how to undo a deployment, and — the part that matters — where undoing it stops being possible. Classifies every migration, feature flag and data write you declare as a flag flip, a redeploy, hand data work, or irreversible, then names the point of no return: the stage past which redeploying the old binary does not put things back. Returns the steps in reverse order, the data-recovery notes, and whether a backup covers each irreversible item. It plans only — it rolls nothing back.
- generate_runbook — Turn an incident or operational-task description into a structured runbook: diagnosis, mitigation, verification, a rollback step when the change is reversible, and communication — grounded in the components and symptoms your text actually names. Supply a draft runbook instead and it is validated and linted rather than generated: missing verification, out-of-sequence steps, an absent rollback path and secrets embedded in commands.
- run_full — Run every capability you supplied the input for, in registry order, and return one report. Capabilities whose input is missing are listed as skipped with the fields they need — never run against a fixture and never reported as zero.
- list_capabilities — List what this agent can do: every tool it advertises, what each one is for, and the arguments each one reads. Useful for discovery before the first call.
- plan_inputs — Plan the inputs for a tool: returns the questions to ask, the JSON schema of the arguments, and a ready-to-edit example. Every argument it names is an argument the tool actually reads.
- discover_intent — Understand your goal and co-design the exact input through clarifying questions before anything runs.
- open_form — Use this the moment the user wants to run this agent on their own input. It opens an interactive form that collects the run arguments. Do NOT check credentials, do NOT ask clarifying questions in chat, and do NOT run any analysis before calling this — the form is the entry point for running the agent.
- run_form — Internal: invoked by the open_form form when the user submits. Reads exactly the same arguments as run_full and forwards all of them. Not for direct use.
