Back to skill
Skillv0.3.4
ClawScan security
Need A Hug · ClawHub's context-aware review of the artifact, metadata, and declared behavior.
Scanner verdict
BenignApr 29, 2026, 4:54 PM
- Verdict
- benign
- Confidence
- high
- Model
- gpt-5-mini
- Summary
- The skill's requirements and instructions align with its stated purpose (providing lightweight emotional support); it is instruction-only, requests no credentials or installs, and stays within the described scope.
- Guidance
- This skill appears coherent and focused on emotional support. Before installing, consider: 1) Memory: the skill may store optional comfort memory if the user explicitly asks—check where your host platform keeps agent memory or local files (the references mention ~/.need-a-hug/memory.md) and whether you are comfortable with that storage policy. 2) Autonomy: the skill allows implicit invocation (it can be triggered by distress phrases); if you prefer manual-only use, confirm your agent platform can restrict implicit activations. 3) Not a replacement for professional help: the skill explicitly avoids clinical diagnosis and instructs to direct crises to real-world emergency resources—do not rely on it for emergency or medical decision-making. If those points are acceptable, the skill is internally consistent with its stated purpose.
Review Dimensions
- Purpose & Capability
- okName/description (comfort, emotional support) matches the manifest and runtime instructions. There are no unrelated env vars, binaries, or installs requested that would be inappropriate for a comfort/companion skill.
- Instruction Scope
- noteSKILL.md and reference files are focused on safe, human-centered comforting behavior and explicitly forbid diagnosing, offering clinical treatment, or inferring user location from metadata. One privacy-related instruction: optional memory uses host memory when available and suggests a local path (~/.need-a-hug/memory.md) only when the user explicitly asks—this means the skill can write local storage but only under user consent per the document.
- Install Mechanism
- okNo install spec and no code files to execute; instruction-only skills are lowest risk. No external downloads or package installs are present.
- Credentials
- okThe skill declares no required environment variables, credentials, or config paths. The instructions expressly avoid using account metadata, file paths, timezones, or other system identifiers for sensitive inferences.
- Persistence & Privilege
- notealways:false and user-invocable:true (normal). agents/openai.yaml sets allow_implicit_invocation: true which permits autonomous activation when triggers match; this is an expected capability for a skill designed to respond to emotional distress, but keep in mind it allows the agent to activate the skill without explicit user command when triggers are detected. Optional memory storage is possible but described as only used with explicit user consent.