founder-playbook-skill
Pass
Audited by Gen Agent Trust Hub on Jul 30, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The bundled script
scripts/analyze_recording.pyusessubprocess.runto invokeffmpegandffprobefor audio track extraction and metadata probing. These invocations use argument lists rather than raw shell strings, which is a secure practice that prevents command injection. - [EXTERNAL_DOWNLOADS]: The skill instructions and metadata recommend installing several standard third-party Python libraries (such as
numpy,scipy,librosa,pyloudnorm,pesq,pystoi, andsoundfile) and theffmpegsystem binary from official public registries (PyPI and Homebrew). - [PROMPT_INJECTION]: An indirect prompt injection surface is present because the analysis script reads and outputs metadata fields (e.g.,
title,appName,speakers) frommetadata.jsonand audio track information provided in the input directory. Maliciously crafted metadata in files from an untrusted source could be used to influence the AI agent's behavior during the analysis phase. - Ingestion points: Metadata and track information from the user-specified recording directory in
scripts/analyze_recording.py. - Boundary markers: None present.
- Capability inventory: Local file system access and subprocess execution for audio analysis and processing.
- Sanitization: Metadata strings are extracted and printed to the console without sanitization or escaping.
Audit Metadata