verdict-dfir

Drive a coding task to a verified, 5-star finish with a bounded agentic OODA loop (observe, orient, plan, act) gated by TDD, an explicit Definition of Done, and a code-quality bar, then commit and open a PR to the dev remote. Runs in an isolated git worktree and keeps a durable plan/checklist on disk so an interrupted run can resume. Use this whenever the user wants something built, fixed, or refactored "until it's done", "until tests pass", or "until it meets the definition of done", asks to iterate/loop on an implementation, wants 5-star repo/code quality, or says run dev-loop / agentic loop / OODA loop / TDD loop. Prefer this over a one-shot edit whenever the first attempt is unlikely to be the final answer (bugfixes, flaky tests, coverage gaps, multi-step features) and the work should end in a real green build and a PR, not a claim.

Run VERDICT DFIR end to end from Claude Code. Use when the operator says /verdict, run verdict, investigate evidence end to end, is there evil here, or wants a signed Verdict and report. The skill verifies setup, uses SIFT mode when available, falls back honestly when it is not, and reports dashboard/report paths without treating optional automation as evidence.