The Hidden Drop In SHiPlog Hunt Reports

by Jule 40 views
The Hidden Drop In SHiPlog Hunt Reports

Running /shiplog hunt today, you might think all signed review comments are counted - yet the system still flags six open PRs as unreviewed, even though review artifacts live deep in PR comments. A March 2026 audit reveals a quiet but costly disconnect: the hunt command relies on outdated logic that misses stale review signals, while installed copies or cached command versions mislead triage reports. This isn’t just a bug - it’s a misalignment between how review state is tracked and how it’s reported.

The current hunting model prioritizes formal GitHub review flags over the baked-in Reviewed-by: comments in PRs. Signed contributions from Codex passes go unrecognized, and the command fails to pull true review provenance. This creates confusion, especially when a PR was reviewed via a signed comment but still shows ‘no reviews’ - a gap that invites duplicate checks and wasted effort.

Culturally, this mirrors a broader US trend: teams trust formal APIs over granular artifact logs, even when the logs contain the full audit trail. On TikTok, viewers mock hunt commands that misfire; in Slack channels, engineers vent about redundant reviews. But beneath the friction lies a clear pattern: the hunt command’s loading path - whether from repo, installed skill, or cached script - determines what gets counted.

  • Stale installed command copies can override dynamic review state logic.
  • Different deployments load shiplog via distinct entry points, leading to inconsistent output.
  • Claude’s output depends on which version of the command is active - sometimes the old one, sometimes the new.

The elephant in the room? A 2026 change proposal for identity-aware hunting remains unmerged, so the old mismatch persists. Fixing it demands more than code: it requires aligning tooling, docs, and user expectations.

Modern review workflows demand precision. When did a signed comment stop being treated as a primary signal? How many hours are lost chasing phantom reviews? The hunt command isn’t broken - it’s outdated. A tier-2 audit confirms: fix the loading paths, honor signed artifacts, and redefine how review state is detected. The next hunt should not just check PRs - it should truly review them.

Is your hunt reporting what’s real? Or just echoing a relic?