← ClaudeAtlas

rig-debuglisted

Systematic root-cause debugging for a failing test, production bug, flaky behavior, or unexplained error. Spawns the debugger agent through Phase 1 → 4 (root-cause → pattern analysis → hypothesis → minimal fix). Refuses to propose fixes before evidence is gathered. Triggers on: 'debug', 'why is X failing', 'root cause', 'this isn't working'.
agent-rig/rig · ★ 3 · Code & Development · score 70
Install: claude install-skill agent-rig/rig
# Debug Drive a bug to root cause and a single minimal fix using the four-phase methodology. ## Configuration Reads `.rig/config.json`: - `test.command` — how to run the test suite when collecting Phase 1 evidence (default: `npm test`). - `agents.debugger` — the project's name for the canonical `debugger` role (default: `debugger`). - `agents.architect` — the project's name for the canonical `architect` role, used only when escalating (default: `architect`). If the file is absent, use the defaults above and note you're running unconfigured. ## The four-phase methodology The `debugger` agent runs this loop; the skill drives it and refuses to let it skip ahead. Each phase has a deliverable that gates the next: - **Phase 1 — Root cause.** Gather evidence: reproduce the failure, read the failing code path, capture the exact error/stack/trace, `git diff` the suspect window. Deliverable: a stated root cause backed by observed evidence — not a guess. - **Phase 2 — Pattern analysis.** Ask whether this is an instance of a class. Does the same bug shape exist elsewhere in the tree? Deliverable: the blast radius (this one site, or N sites). - **Phase 3 — Hypothesis.** State the specific change that should fix the root cause and why, plus how you'll confirm it. Deliverable: one falsifiable hypothesis. - **Phase 4 — Minimal fix.** Apply the smallest change that addresses the root cause, add/extend a test that fails before and passes after. Deliverable: root