What problem does it solve? When a stack trace, failing test, or production alert appears mid-task, engineers often either bypass their workflow entirely with an untracked hotfix or over-apply full ceremony to a one-line fix. This Skill triages concrete failure evidence into a structured classification and routes it to the smallest decisive next step, keeping urgent incidents auditable without unnecessary process. ## Core Features & Use Cases - Failure classification: Categorizes the incident into exactly one of REGRESSION, NEW_BUG, CONFIG, EXTERNAL_DEPENDENCY, REPRODUCIBILITY, or DIAGNOSTIC_INSUFFICIENT based on the pasted failure signal. - Fix-size routing: Recommends HOTFIX, SLICE, INVESTIGATION, or ESCALATE and routes to the matching follow-up command (branch-commit, implement-approved-slice, impact-analysis, or capture-observation). - Decision and invariant validation: Checks the proposed fix path against locked decisions in DECISIONS.md and INVARIANTS_AND_NON_GOALS.md, surfacing conflicts instead of silently overriding them. - Use Case: A CI run fails with a stack trace after a recent deploy. Paste the trace, expected behavior, environment, and urgency tag; the Skill classifies it as a REGRESSION, recommends a HOTFIX with a safety justification, and records the decision in the task folder. ## Quick Start Triage this failing test output against my active task folder and tell me whether it needs a hotfix or a full slice.
