hyperframe-pro-beats
Pass
Audited by Gen Agent Trust Hub on Sep 22, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute a local Python script located at
${CLAUDE_PLUGIN_ROOT}/scripts/shots.pyusingpython3. This script is utilized to find, fetch, and process screenshots based on external URLs provided as arguments during the video planning phase.\n- [EXTERNAL_DOWNLOADS]: The skill facilitates the download of images from arbitrary remote URLs via thefetchcommand in its internal script. This allows external content to be brought into the agent's execution environment from sources that are not pre-whitelisted.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from the web, creating a vulnerability surface for indirect prompt injection. Malicious instructions could potentially be embedded in external website text or image metadata to influence the agent's subsequent actions.\n - Ingestion points: External URLs used to find and fetch images (
find \"<article-url>\",fetch \"<img-url>\") and their associated web content.\n - Boundary markers: The skill does not define specific automated boundary markers for the data ingested, relying instead on the agent's review and manual steps.\n
- Capability inventory: The skill utilizes subprocess execution (
python3), network requests for file fetching, and local file system writes (raw/a.jpg,out.png).\n - Sanitization: No automated sanitization is provided for the fetched content; however, the skill provides manual safety guidance to redact sensitive information from screenshots before use.
Audit Metadata