← ClaudeAtlas

error-fixlisted

Read error-report.md, fetch each New/Regressed issue's Sentry detail and stack trace, classify hitl per issue, and queue task-store tasks (plus companion observability-fix tasks where instrumentation gaps hinder root-causing). Requires error-scan to have run first.
app-vitals/shipwright · ★ 6 · Code & Development · score 71
Install: claude install-skill app-vitals/shipwright
# Error Fix Read the latest `error-report.md` and, for every unresolved New/Regressed issue that has a mapped repo, fetch its Sentry detail and latest-event stack trace, classify whether it can be fixed autonomously or needs a human, and queue it as a task-store task. Findings are never turned into direct PRs; they always become task-store tasks that `dev-task` (or a human, for HITL tasks) picks up later. Issues whose service tag is `UNMAPPED` in the report are never queued — they're surfaced for human triage instead, since guessing a repo could send a fix into the wrong codebase. **Prerequisites:** - Run `/error-scan` first to produce `error-report.md`. - `SENTRY_ORG` and `SENTRY_AUTH_TOKEN` must be set in the environment — this skill calls the Sentry API directly (issue detail + latest event) to gather enough context for HITL classification and root-cause investigation, the same env vars `/error-scan` requires. > **Task store setup:** This skill pushes findings to the Shipwright task store. If > `SHIPWRIGHT_TASK_STORE_URL` or `SHIPWRIGHT_TASK_STORE_TOKEN` is missing, invoke > `/shipwright:task-store` for setup instructions. --- ## Setup: Parse Arguments Before starting, check for flags: - `--dry-run` — print what would be queued (including hitl classification and companion tasks) without querying the task store for dedup and without writing anything (no task-store writes, no ledger writes) > **Note:** Queueing is the only mode. There is no PR mode and no `-