Menu
Optional defect-repair profile for bug search, regression repair, flaky failure investigation, incidents, and security-bug fixes.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Based on SOC occupation classification
Create or revise production-ready SDD plans for agent execution, including clarification, specification, freeze inputs, ownership, budgets, and evidence contracts.
Independently audit SDD specifications and agent-ready plans before freeze, returning findings, readiness verdicts, and required refinements.
Independently audit a task attempt or completed run against a frozen plan, checking ownership, evidence, acceptance, and final readiness.
Convert external issue payloads into draft-only ALK specification input without starting execution or bypassing review/freeze.
Compile a frozen reviewed plan into deterministic worker packets with dependency waves, ownership, validation, evidence, and result contracts.
Run or resume the full lifecycle from request clarification through reviewed planning, freeze, authorized execution, per-task audits, and final proof.
| name | bug-forensics |
| description | Optional defect-repair profile for bug search, regression repair, flaky failure investigation, incidents, and security-bug fixes. |
Use this skill only when the task explicitly asks to find or repair a bug, regression, flaky failure, incident, or security bug. The profile is optional and disabled by default for ordinary feature work.
Before accepting a patch, collect schema-backed evidence for:
Required phase-1 receipts:
agent-bug-reproduction-receipt.v1;agent-failure-fingerprint.v1;agent-bug-hypothesis-ledger.v1;agent-regression-proof-receipt.v1;agent-fix-impact-receipt.v1.Reuse agent-fix-impact-receipt.v1 for behavior impact. Do not create a
bug-specific competing fix-impact schema.
agent-cross-check-receipt.v1 with token/resource
caps.For compact hosts, include only the active bug packet, reproduction command, fingerprint fields, current hypotheses, suspect/write scope, and validation receipts. Summarize logs by digest and short failure pattern instead of copying large traces.