Skillv1.0.0

ClawScan security

defuddle-web-cleaner · ClawHub's context-aware review of the artifact, metadata, and declared behavior.

Scanner verdict

ReviewMar 14, 2026, 12:18 PM

Verdict: Review
Confidence: medium
Model: gpt-5-mini
Summary: The skill's instructions rely on an external 'Defuddle' parser/CLI but the package declares no install requirements or required binary—this mismatch makes the skill's runtime behavior unclear and potentially unsafe.
Guidance: This skill appears to be an instruction-only wrapper that expects a separate 'defuddle' parser/CLI, but it doesn't declare how that tool is provided. Before installing or enabling: 1) Confirm whether your agent environment already has a trusted 'defuddle' binary/service available; ask the author for the install steps and the exact binary or endpoint used. 2) If the skill will download or run a binary at runtime, request the download URL, checksum, and source (official release page) and avoid unknown personal servers. 3) Test the skill on non-sensitive, public pages first to observe network activity and outputs. 4) Consider privacy: the agent will fetch URLs (revealing your IP) and process page content—do not send pages containing secrets or private auth tokens. 5) If you need stronger assurance, ask the publisher for source code or a vetted install spec; until then treat the skill as potentially unsafe.

Review Dimensions

Purpose & Capability: concernThe SKILL.md repeatedly refers to running a 'Defuddle parser' and shows example CLI and curl usage, but the registry metadata declares no required binaries, no install steps, and ships no code. Either the environment must already have 'defuddle' available (not stated), or the skill will need to fetch/install a binary at runtime. That omission is disproportionate to the stated purpose and creates ambiguity about what will actually run.
Instruction Scope: noteInstructions are high-level and stay within the expected domain (fetch page HTML, parse, extract metadata, convert to markdown). However they are vague about how to 'Load page HTML' and how to 'Run Defuddle parser', leaving the agent broad discretion to choose network/fetch methods or to download/run tooling. That vagueness increases risk because the agent may perform unexpected network access or execute unsigned code.
Install Mechanism: concernThere is no install specification and no shipped code. The references imply a CLI/service named 'defuddle' (examples: 'defuddle parse', 'curl defuddle.md/...'), but the skill doesn't declare that the binary is required or how to obtain it. This missing install info is a red flag: if the agent attempts to acquire and run a binary at runtime, it could pull arbitrary code without review.
Credentials: okThe skill does not request any environment variables, credentials, or config paths, which is proportionate for a web-cleaning utility. There is no evidence it asks for unrelated secrets. The main concern is the unspecified dependency on an external tool rather than overbroad credential requests.
Persistence & Privilege: okThe skill does not request persistent/always-on privileges and is user-invocable only. It does not declare modifications to other skills or system-wide settings—no privilege escalation is apparent from the metadata.