Back to skill
Skillv1.0.0
ClawScan security
Honest Critic · ClawHub's context-aware review of the artifact, metadata, and declared behavior.
Scanner verdict
BenignApr 29, 2026, 4:48 AM
- Verdict
- benign
- Confidence
- high
- Model
- gpt-5-mini
- Summary
- The skill is an instruction-only “honest critic” prompt that asks Claude to push back and provide prioritized critiques; it requests no credentials or installs and its behavior matches its description.
- Guidance
- This skill is internally consistent and doesn't ask for credentials or installs. Before enabling it widely: (1) test it on non-critical prompts to see whether its blunt ‘‘Verdict’’ style gives useful (not overconfident) answers for your workflows; (2) consider disabling autonomous invocation or only calling it when you explicitly request a tough critique; and (3) pair it with a clarity prompt (define scope and constraints) so the critic evaluates the right thing and doesn't overcommit when nuance is required.
Review Dimensions
- Purpose & Capability
- okName, description, and runtime instructions are aligned: the SKILL.md defines a structured critique protocol (red-flag scan, steelman, prioritized pushback, verdict) and does not ask for unrelated capabilities, binaries, or credentials.
- Instruction Scope
- noteInstructions are narrowly scoped to evaluating user-provided ideas and specify a clear output format. One practical risk: the directive to "Never use 'it depends' as a final answer" may encourage overconfident, under‑nuanced recommendations in cases that truly require caveats; otherwise the instructions do not request system files, env vars, or external endpoints.
- Install Mechanism
- okNo install spec or code is present (instruction-only skill), so nothing is written to disk or downloaded. The included example 'openclaw install' commands are user-facing convenience suggestions but are not an install script in the SKILL.md.
- Credentials
- okThe skill declares no required environment variables, credentials, or config paths. There is no disproportionate access requested for its stated purpose.
- Persistence & Privilege
- noteFlags are default (always: false, user-invocable: true, disable-model-invocation: false). Autonomous invocation of the skill by the agent is permitted (this is the platform default). Because the skill encourages definitive verdicts, consider whether autonomous use without user confirmation is desirable in your context.