Back to skill
Skillv1.0.0

ClawScan security

Head Tool · ClawHub's context-aware review of the artifact, metadata, and declared behavior.

Scanner verdict

SuspiciousApr 28, 2026, 6:15 PM
Verdict
suspicious
Confidence
medium
Model
gpt-5-mini
Summary
The skill is a simple file-preview tool and contains no network/credential behavior, but the documentation (SKILL.md) and declared usage claim features that the included script does not implement, so the package is internally inconsistent.
Guidance
This skill appears to be what it claims (a simple head-like viewer) and contains no network or credential requests, but the documentation and examples promise features (-c byte mode, -q quiet, multiple files) that the included script does not implement. Before installing or relying on it: 1) treat the package as untrusted code and review the script (it's short) or run it in a sandbox; 2) be aware the script reads entire files into memory (open(...).readlines()), which can cause large-memory use on very big files — prefer a streaming implementation if you need to inspect huge logs; 3) if you need the advertised -c/-q/multiple-file behavior, either update the code or do not rely on this skill; 4) since the skill can read any file the agent has permission to access, avoid invoking it on sensitive paths unless you trust the environment. If you want a fully coherent skill, ask the author to either update SKILL.md to match the code or extend the script to implement the advertised options and safe streaming/encoding handling.

Review Dimensions

Purpose & Capability
noteThe stated purpose (show first lines of files) matches the supplied script: scripts/head.py prints the first N lines or reads from stdin. However SKILL.md advertises additional options (-c to show bytes, -q quiet mode, multiple files) that the script does not implement; this mismatch is unexplained and could confuse users or hide missing functionality.
Instruction Scope
noteSKILL.md instructs the agent to run head-tool and to read from stdin when no file is specified, which the script follows. The instructions imply handling multiple files and byte-mode output, but the runtime code only accepts a single optional filename and a line-count flag, so the instruction set is broader than the actual code behavior.
Install Mechanism
okThis is an instruction-only skill with one small Python script and no install spec; nothing is downloaded or installed automatically, which minimizes risk.
Credentials
okNo environment variables, credentials, or config paths are requested or required; the script only reads files or stdin, which is consistent with purpose.
Persistence & Privilege
okThe skill does not request always-on presence and has normal invocable defaults; it does not modify system or other skills' configs.