bat-submit
Pass
Audited by Gen Agent Trust Hub on Aug 14, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [PROMPT_INJECTION]: Indirect Prompt Injection Attack Surface\n
- Ingestion points: External website content is ingested via the bat-cli crawl functionality and the scripts/extract_video.py script (SKILL.md, references/01-extract.md).\n
- Boundary markers: Absent. The instructions do not specify delimiters or provide warnings for the agent to ignore instructions embedded in the processed external content.\n
- Capability inventory: The skill uses bat-cli to write local files (base.json, en.json) and perform network submissions to the vendor API (references/01-extract.md, references/03-submit.md).\n
- Sanitization: Absent. There is no instruction to sanitize, escape, or validate the external content before it is processed or submitted.\n- [COMMAND_EXECUTION]: Autonomous Execution Pattern. The skill explicitly directs the agent to execute the workflow in three sequential steps without pausing for user confirmation (SKILL.md). This autonomy reduces the opportunity for a human to review the results of the extraction and translation phases before final submission, which could be exploited via indirect prompt injection.\n- [COMMAND_EXECUTION]: CLI and Tool Usage. The skill uses npm to install @bataitools/bat-cli and executes various bat-cli subcommands to manage the submission lifecycle (SKILL.md, references/03-submit.md).\n- [EXTERNAL_DOWNLOADS]: Remote Content Retrieval. The scripts/extract_video.py script utilizes the Python urllib library to fetch HTML content from remote URLs for metadata extraction (scripts/extract_video.py).
Audit Metadata