interrogate
Review a concrete diff or implementation for correctness, spec compliance, and codebase fit.
Skill by grugh
Install
Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.
npx agentmarket install skill/grugh-interrogate
Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.
Get the weekly diff
The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.
About
Use cases: documents, security
Reviews
Write a reviewNo reviews yet. Installed it? Reviews come from verified installers only. How reviews work
Permissions
What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.
- Tools
- No tools found
- Secrets
- None asked for
- Install scripts
- None found
Network hosts it talks to 1
codeload.github.com· found in install command
Files it touches
~/.claude/skills/interrogate
What's inside
- Skill name
interrogate- Description
- Review a concrete diff or implementation for correctness, spec compliance, and codebase fit. Report evidence-backed findings without automatically fixing code.
- License
- MIT
- Metadata
- source: Lauren Tan, pstack: interrogate; Matt Pocock, skills: code-review. · opencode/autoinvoke: false
- Instructions
- 25 lines in SKILL.md · view on GitHub
Files (9)
- LICENSE
- SKILL.md
- openai.yaml · agents/
- adversarial.md · references/
- code-quality-review.md · references/
- lead-judgment.md · references/
- reviewer-prompt.md · references/
- rubric.md · references/
- standards-and-spec.md · references/
SKILL.md (first lines)
# Interrogate Review without automatically fixing code or posting externally. Preserve the two source methods as distinct modes: - **Adversarial** (default): correctness, security, root causes, structural quality, and verification. Follow [the adversarial review](references/adversarial.md), including all four bundled rubric and judgment references. - **Standards and spec**: documented repository standards and requirement compliance. Follow [the standards/spec review](references/standards-and-spec.md), including its full smell baseline and separate reports. - **Both**: run both methods over the same scope. Report the adversarial verdict first, then the Standards and Spec reports. Do not merge or rerank findings across these methods or across the Standards and Spec axes. Cross-reference repeated evidence without hiding an axis's finding. Use the mode the user names. A request specifically about standards/spec compliance selects that mode; a request for both methods or a full review selects Both; anything else, including a general review, uses Adversarial. State the selected mode before reviewing. These are procedures within this skill, not calls to other skills. ## Establish a shared scope For a branch or committed-change review, resolve the intended fixed point and verify the comparison before reviewing. Ask if it is missing or ambiguous; do not assume a branch name. Use `git diff <fixed-point>...HEAD` and `git log <fixed-point>..HEAD --oneline`. Report an invalid reference or empty scope before launching reviewers. For an explicit working-tree review, include staged and unstaged changes and relevant untracked files; do not substitute a HEAD-only diff that omits the requested work. Record the exact scope and use it for every selected mode. For an explicit file review, identify the files and surrounding callers needed to assess them. These explicit scopes replace the fixed-point requirement in the standards/spec procedure. Identify intended behavior from the request, requirements, and relevant history. Read the actual specification where available; code may show behavior but cannot establish requirement compliance by itself. Give reviewers the scope, relevant context, and requirements, not your conclusions. ## Execution and evidence Use independent contexts and model diversity when available and authorized. Sequential fallback preserves the rubrics and output contracts but is not an independent review; disclose the difference. Every adversarial reviewer receives the whole rubric. Standards and Spec remain separate reviews. Verify reported execution paths and citations before returning the review. Run focused checks when useful and permitted and record what actually ran and what remains unverified. Do not change source, snapshots, or tests as part of the review. Zero findings is valid; state coverage limits.
From grugh/skills, MIT.
More from grugh/skills
- 0.567score
blast-radius
grugh
Trace what outside a proposed or implemented change could break, especially shared contracts, schemas, lifecycle, serialization, concurrency, and reused…
SkillAFreshness: AClaude Code +9 - 0.567score
architecture-review
grugh
Audit an existing repository or subsystem for high-value structural simplifications, clearer ownership, and better boundaries.
SkillAFreshness: AClaude Code +9 - 0.567score
how
grugh
Explain how an existing subsystem, feature, or code path works now by tracing repository evidence.
SkillAFreshness: AClaude Code +9 - 0.567score
why
grugh
Investigate the historical rationale behind code or architecture using source, Git history, and available discussions.
SkillAFreshness: AClaude Code +9 - 0.565score
grug
grugh
Challenge unnecessary complexity in a design, implementation, or technical decision when the user explicitly asks for Grug's perspective.
SkillAFreshness: AClaude Code +9 - 0.561score
remember
grugh
Remember a task across sessions in project-local state.
SkillAFreshness: AClaude Code +9 - 0.561score
technical-writing
grugh
Layered technical-writing standard: Diátaxis structure, Google developer style sentences, STE instruction rules, Global English syntax.
SkillAFreshness: AClaude Code +9 - 0.559score
innovate
grugh
Propose the single most valuable and inventive addition to the current project when the user asks for a fresh improvement idea.
SkillAFreshness: AClaude Code +9 - 0.559score
caveman
grugh
Use playful caveman-style speech when the user explicitly requests caveman mode.
SkillAFreshness: AClaude Code +9 - 0.539score
grilling
grugh
Grill the user relentlessly about a plan, decision, or idea.
SkillAFreshness: AClaude Code +9 - 0.533score
unslop
grugh
Cut AI tells from any writing.
SkillAFreshness: AClaude Code +9 - 0.525score
restate
grugh
Check shared understanding of the user's intent.
SkillAFreshness: AClaude Code +9
Signals
- Stars
- 0
- Forks
- 0
- Open issues
- 0
- Last commit (this folder)
- Oct 13, 2026 (16 days ago)
From GitHub (fetched Oct 24, 2026).
Details
- License
- MIT
- Repository
- github.com/grugh/skills
- First seen
- Oct 24, 2026
- Last checked
- Oct 24, 2026
- Source
- GitHubgithub-graphql
More from grugh
- 0.567score
blast-radius
grugh
Trace what outside a proposed or implemented change could break, especially shared contracts, schemas, lifecycle, serialization, concurrency, and reused…
SkillAFreshness: AClaude Code +9 - 0.567score
architecture-review
grugh
Audit an existing repository or subsystem for high-value structural simplifications, clearer ownership, and better boundaries.
SkillAFreshness: AClaude Code +9 - 0.567score
how
grugh
Explain how an existing subsystem, feature, or code path works now by tracing repository evidence.
SkillAFreshness: AClaude Code +9 - 0.567score
why
grugh
Investigate the historical rationale behind code or architecture using source, Git history, and available discussions.
SkillAFreshness: AClaude Code +9 - 0.565score
grug
grugh
Challenge unnecessary complexity in a design, implementation, or technical decision when the user explicitly asks for Grug's perspective.
SkillAFreshness: AClaude Code +9 - 0.561score
remember
grugh
Remember a task across sessions in project-local state.
SkillAFreshness: AClaude Code +9
Cite this page
interrogate skill. Agent Mart. https://agentmart.dev/skills/grugh-interrogate (last updated 2026-10-24)