Back to skill
Skillv1.0.0
ClawScan security
AANA File Operation Guardrail Skill · ClawHub's context-aware review of the artifact, metadata, and declared behavior.
Scanner verdict
BenignMay 2, 2026, 9:41 PM
- Verdict
- Benign
- Confidence
- high
- Model
- gpt-5.5
- Summary
- This is a coherent instruction-only safety guardrail for file operations and does not include code, commands, credentials, persistence, or hidden data flows.
- Guidance
- This skill appears safe to install as a file-operation review checklist. It should make agents more cautious before deleting, overwriting, moving, publishing, or bulk-editing files, but you should still read any approval prompt carefully and confirm the exact target paths before allowing irreversible actions.
Review Dimensions
- Purpose & Capability
- okThe stated purpose is to make risky file operations safer, and the instructions consistently require scoping, reversibility, and explicit approval before destructive, broad, or publishing actions.
- Instruction Scope
- okThe skill limits agent behavior rather than expanding it: it tells the agent to ask, defer, revise, or refuse when file scope, ownership, authorization, or reversibility is unclear.
- Install Mechanism
- okThere is no install spec, no bundled code, no dependencies, and no command execution. The manifest and README both describe it as instruction-only.
- Credentials
- okThe artifacts do not request environment variables, credentials, filesystem inspection, network access, or OS-specific authority. Listed capability signals are not supported by executable artifacts.
- Persistence & Privilege
- okThe skill explicitly says it does not persist memory and instructs agents not to store file contents or paths without permission.