promo-video
Fail
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: HIGHINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill defines a workspace analysis feature in
SKILL.mdthat ingests untrusted repository content to automatically customize video creation parameters. - Ingestion points: The agent runs
git log --oneline -100and reads local files (README.md,app/models/) to extract project features and context. - Boundary markers: There are no clear boundary markers or instructions to ignore embedded prompts within the files being scanned.
- Capability inventory: The skill has extensive execution capabilities including project creation via
npx, dependency installation vianpm, and multimedia processing viaffmpegandpythontools. - Sanitization: No sanitization or escaping mechanisms are applied to data pulled from the git log or local documentation before it is processed by the agent.
- [COMMAND_EXECUTION]: The voiceover script (
scripts/generate_voiceover.py) usessubprocess.runto call system utilities (ffmpegandffprobe) to perform essential video production tasks, including calculating audio length, generating silence loops, and blending background audio. - [EXTERNAL_DOWNLOADS]: The documentation (
README.md) provides setup guidance directing users to manually download dependencies, including referencing the official Homebrew installer shell script on GitHub. This is a standard tech platform installation method and poses no direct automated execution risk.
Recommendations
- HIGH: Downloads and executes remote code from: https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh, unknown (check file) - DO NOT USE without thorough review
Audit Metadata