repo-evidence-first

Grounding code review claims in repository evidence requires classifying each assertion as a Grounded Fact, Hypothesis, or Unverified state based on inspected file anchors. This structured evidence pack workflow prevents confident but unsupported statements during non-trivial reviews.

Checking test coverage for completion claims involves creating a structured evidence pack that explicitly verifies test paths and coverage assertions. This process classifies claims as Grounded Fact or Open Question based on inspected repository anchors before stating the outcome.

Detecting stale code and resolving file authority conflicts requires constructing a structured evidence pack that evaluates current versus stale status. It routes to lightweight specialists to obtain compact anchors, ensuring non-trivial planning is backed by grounded repository facts.

Repo evidence packs support security assurance and workflow planning by forcing evidence-backed verification for any non-trivial editing or completion assertion. They dispatch to security review or code scouting roles to obtain compact anchors instead of dumping logs.

Repo evidence status verification is limited by dependencies on inspected repository anchors and targeted routing to specialists. Claims remain classified as Blocked or Open Question when authoritative files, test coverage paths, or verification data cannot be explicitly grounded in the repository.

Generating an evidence pack is needed before stating task completion whenever non-trivial claims about current versus stale status, authoritative files, or test coverage are made. It forces structured verification through explicit repo evidence status and anchor-based classification.