What problem does it solve? Manually re-checking a spec, test suite, or document after every fix is slow and error-prone, and it is easy to stop too early or loop forever. This Skill automates a bounded improvement loop: research agents find gaps, fix agents apply corrections, and the loop repeats until no gaps remain, a numeric metric goal is hit, or a plateau/cap/regression halt condition triggers. ## Core Features & Use Cases - Iterative gap-fix loop: Spawns parallel research agents to find FAIL/CONCERN/OBSERVATION gaps, then parallel fix agents to resolve them, with per-iteration reports and a TSV results log. - Metric-goal mode: Accepts a verify: shell command that outputs a number (e.g., lint error count, coverage percent) and loops until the target is reached, with an optional guard: command that must stay green after every fix batch. - Domain presets and subcommands: Ships canonical configs for spec, tests, ux, docs, plan, errors, and harness domains, plus probe, reason, and evals subcommands for persona interrogation, adversarial debate, and TSV trend analysis. - Use Case: Point it at a plan document with domain: plan and it runs validate-fix cycles (PVL), writing one report per iteration and stopping when the validator returns PASS or the 10-cycle cap is hit. ## Quick Start Ask the AI to run vc-autoresearch with domain spec against your requirements document and let it loop until no gaps remain.
