Back to skill
Skillv1.0.3
ClawScan security
Calm Down · ClawHub's context-aware review of the artifact, metadata, and declared behavior.
Scanner verdict
BenignApr 12, 2026, 5:17 AM
- Verdict
- benign
- Confidence
- high
- Model
- gpt-5-mini
- Summary
- The skill's instructions, requirements, and behavior are internally consistent with its stated purpose (appending short grounding reminders when conversational frustration is detected); it requests no credentials or installs and contains no code, so there is no sign of hidden data access or unrelated capabilities.
- Guidance
- This skill appears coherent and low-risk, but double-check a few practical things before enabling it broadly: 1) Confirm how the platform provides message timestamps, timezone, and language detection so the skill's 'late-night' and language-trigger rules work correctly. 2) Ask the developer how it avoids repeating the same grounding line twice (session state? ephemeral memory?) and where that state would be stored. 3) Consider user experience: automated reminders can feel patronizing if misfired — ensure you can opt out or tune sensitivity, and test the trigger rules on representative conversations (including polite corrections and external-frustration cases) to avoid false positives. 4) Because it alters responses automatically, limit deployment to agents or users who consent to automated behavioral nudges.
Review Dimensions
- Purpose & Capability
- okName and description (detect frustration + append grounding reminder) match the SKILL.md content. The skill requires no binaries, env vars, or installs that would be unrelated to its purpose.
- Instruction Scope
- noteInstructions are narrowly scoped to detecting conversational signals and appending a single reminder after addressing the user's problem. Two minor implementation notes: (1) rules like 'never repeat the same line twice in a row' imply the skill may need short-term state (to avoid immediate repetition) but no storage or memory mechanism is declared; (2) triggers that depend on 'local time' and 'detect user's language' assume the agent has access to message timestamps and language-detection capability. These are reasonable but worth confirming with the platform implementer.
- Install Mechanism
- okInstruction-only skill with no install spec and no code files — nothing is downloaded or written to disk. Lowest install risk.
- Credentials
- okNo environment variables, credentials, or config paths are requested. All declared requirements are proportional to the purpose.
- Persistence & Privilege
- noteSkill is not set to always: true and does not request elevated persistence. The 'never repeat twice' rule suggests the skill might benefit from remembering the last reminder between invocations; the SKILL.md does not declare where that state would be stored. Confirm whether the platform will supply per-session state or if the skill will require persistent storage.