What problem does it solve? When work stalls, it is hard to see exactly what is stuck, what is missing, and whether status claims rest on real evidence. This Skill answers "what is stuck or missing?" on any project, repo, plan, or task by computing a fresh, read-only assessment of blockers, unresolved decisions, uncommitted work, and inferred silent gaps. ## Core Features & Use Cases - Stall surface detection: Delegates evidence gathering to a bounded, read-only subagent that checks git state (uncommitted files, unmerged branches), plan documents (unchecked items, TODO/FIXME markers), and conversation threads to attribute blockers as on-you versus on-others. - Silent gap inference: Surfaces what is plausibly missing but unproven (no tests next to new code, stale docs), always marked with provenance tiers (attested, inferred, guessed) and never presented as confirmed fact. - Audit mode: The --audit flag raises claim-versus-evidence mismatches and gamed signals, such as a "tests pass" claim with no test files or commit counts padded by whitespace-only changes. - Use Case: Before trusting a status update on a repo, run the gaps check to see that one blocker awaits review, three files are uncommitted, and a schema decision was never recorded, each with a re-runnable source reference. ## Quick Start Ask the agent to run /orient-gaps on the current repository to show what is stuck, what is missing, and whether the evidence supports trusting it.
