Skillv1.0.0

ClawScan security

voice-text-to-meme · ClawHub's context-aware review of the artifact, metadata, and declared behavior.

Scanner verdict

ReviewMar 14, 2026, 9:12 AM

Verdict: Review
Confidence: high
Model: gpt-5-mini
Summary: The skill's functionality (turn voice/text into a meme image) is coherent, but there are inconsistencies around required secrets and installation that the user should review before installing.
Guidance: What to check before installing: - This skill sends user text/prompts and an API key to an external image model (default base URL https://models.audiozen.cn/v1). Verify that provider and endpoint are trustworthy and acceptable for your data/privacy needs. - The script requires MEME_MODEL_API_KEY (a secret). The skill registry metadata does not declare this required env var — do not paste a high-privilege or reused API key; create a dedicated, scoped token if possible. - The package includes requirements.txt (openai, requests) but no install spec; ensure the runtime will install these dependencies or run the script in a safe environment. - The script prints a JSON payload (including the prompt and base_url) to stdout; logs may therefore contain user text and model details—avoid logging sensitive text or run in an environment with appropriate log controls. - The skill writes image files to meme_outputs or temp directories; check file permissions and cleanup policies. - If you need stronger assurance, ask the maintainer to: (1) declare MEME_MODEL_API_KEY in the registry metadata as a required credential, (2) provide an explicit install spec (or container) to install dependencies, and (3) document the model provider and data handling / retention policy for the default base URL.

Review Dimensions

Purpose & Capability: okName/description match the code: the script takes voice/text, builds a prompt and calls an image-generation model (default doubao-seedream via an OpenAI-compatible client). The external model API key and base URL are logically required for this purpose. However, the registry metadata claims no required env vars while the SKILL.md and script require MEME_MODEL_API_KEY — an inconsistency in declared requirements.
Instruction Scope: noteSKILL.md instructs the agent to call scripts/generate_meme.py and to read MEME_MODEL_API_KEY and optional MEME_MODEL_BASE_URL; those actions are within the stated purpose (sending prompts to an image model). The script prints a JSON payload to stdout that includes the prompt, model name and base_url and the (possibly sensitive) source_text — that could surface user text and the model endpoint in logs/stdout. The skill writes output images to disk (meme_outputs or a temp dir) which is expected.
Install Mechanism: noteThere is no install spec, but a requirements.txt and a runnable Python script are included. That means dependencies (openai, requests) must be installed by the host; the skill package does not declare how to install them. No remote downloads of arbitrary archives or suspicious installers are present.
Credentials: concernThe skill requires a sensitive credential (MEME_MODEL_API_KEY) to call an external image model; that is proportionate to its functionality. However, the registry metadata does not declare any required env vars or a primary credential — the required API key is only documented in SKILL.md and used by the code. Users should treat MEME_MODEL_API_KEY as a secret and avoid reusing high-privilege tokens. MEME_MODEL_BASE_URL defaults to a third-party host (https://models.audiozen.cn/v1), so confirm the provider before sending user content.
Persistence & Privilege: okThe skill does not request always:true, does not modify other skills or global agent settings, and only writes its own output image files. Agent autonomous invocation is allowed by default but is not in itself a new risk here.