메뉴
Adversarial bug finding using 3 isolated agents (Hunter, Skeptic, Referee) to find and verify real bugs with high fidelity. Based on bug-hunt by danpeg (https://github.com/danpeg/bug-hunt).
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
SOC 직업 분류 기준
Use when encountering any bug, test failure, or unexpected behavior, before proposing fixes - four-phase framework with built-in backward tracing for deep-stack failures, ensuring root-cause understanding before implementation
Turn the current session into a coordination thread that routes per-branch implementation work to durable, reusable child agents. Use when the user says 'orchestrator on', wants this session to act as chief-of-staff across branches, or asks to route work without implementing locally.
Test-driven development with red-green-refactor loop. Use when user wants to build features or fix bugs using TDD, mentions "red-green-refactor", wants integration tests, or asks for test-first development.
Adversarial questioning and collaborative shaping in one skill. Two modes: stress-test (challenge a plan/design until gaps surface) and shape (iterate on requirements and solution options until alignment). Use when user says 'grill me', 'stress-test this', 'challenge this plan', 'shape this', 'let's explore options', or wants to pressure-test any design before committing.
Compact the current session into a handoff document for another agent or future session to pick up. Captures mission, repo state, role, scope, locked decisions, dispatch sequence, and a pickup checklist. Studio-aware. Use when pausing work, ending a session, switching contexts, or preparing for a fresh agent to continue.
Initialize a new project with PROJECT.md + ROADMAP.md in .project/, set up jira and studio, and route to the first sprint. Use when starting a new project, onboarding a new codebase, or setting up workflow infrastructure for the first time.
| name | nit |
| description | Adversarial bug finding using 3 isolated agents (Hunter, Skeptic, Referee) to find and verify real bugs with high fidelity. Based on bug-hunt by danpeg (https://github.com/danpeg/bug-hunt). |
| argument-hint | [path/to/scan] |
| disable-model-invocation | true |
Run a 3-agent adversarial nit hunt on your codebase. Each agent runs in isolation.
The scan target is: $ARGUMENTS
If no target was specified, scan the current working directory.
Nit respects a project-level ignore file at .claude/nitignore.md. This file tracks:
Entries use a rule field formatted as category::description (e.g. security::hardcoded-token). Nit matches on file + rule to skip known findings.
A template is available at ${CLAUDE_SKILL_DIR}/references/nitignore-template.md.
You MUST follow these steps in exact order. Each agent runs as a separate subagent via the Agent tool to ensure context isolation.
Read these files using the skill directory variable:
Also check if .claude/nitignore.md exists in the project root. If it does, read it and pass the ignore entries to each agent so they can skip already-triaged findings.
Launch a general-purpose subagent with the hunter prompt. Include the scan target in the agent's task. The Hunter must use tools (Read, Glob, Grep) to examine the actual code.
Wait for the Hunter to complete and capture its full output.
If the Hunter reported TOTAL FINDINGS: 0, skip Steps 3-4 and go directly to Step 5 with a clean report. No need to run Skeptic and Referee on zero findings.
Launch a NEW general-purpose subagent with the skeptic prompt. Inject the Hunter's structured bug list (BUG-IDs, files, lines, claims, evidence, severity, points). Do NOT include any narrative or methodology text outside the structured findings.
The Skeptic must independently read the code to verify each claim.
Wait for the Skeptic to complete and capture its full output.
Launch a NEW general-purpose subagent with the referee prompt. Inject BOTH:
The Referee must independently read the code to make final judgments.
Wait for the Referee to complete and capture its full output.
Display the Referee's final verified nit report to the user. Include:
If zero bugs were confirmed, say so clearly — a clean report is a good result.
After presenting findings, offer the user the option to triage. For each confirmed bug, output a pre-formatted ignore entry the user can approve:
| intentional | category::description | file/path | line(s) | [user adds rationale] |
Ask: "Would you like to dismiss any of these as intentional, or mark any as remediated? I can update .claude/nitignore.md for you."
If the user selects entries to triage:
.claude/nitignore.md doesn't exist, create it from ${CLAUDE_SKILL_DIR}/references/nitignore-template.md