The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill references the installation of protobuf (distributed by a well-known service) and pywidevine, which is a recognized community tool for Widevine L3 research. These are standard dependencies for the stated purpose of DRM analysis.
[COMMAND_EXECUTION]: The instructions utilize standard command-line utilities for forensic analysis, including grep, readelf, tshark, ffprobe, and mp4dump. These tools are used appropriately within the context of security research and binary auditing.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted external data sources, which creates a theoretical surface for indirect prompt injection.
Ingestion points: Untrusted data enters the agent's context through media files (MP4/CMAF), PSSH box metadata, and network captures of license challenge/response traffic as described in SKILL.md (Steps 2 and 3).
Boundary markers: The skill contains extensive warnings and scope limitations, repeatedly emphasizing that analysis must be conducted within "controlled environments" and strictly for "authorized research" purposes.
Capability inventory: Analysis involves local file inspection via grep, binary header auditing with readelf, and media metadata extraction using ffprobe and mp4dump.
Sanitization: The skill employs structured binary parsing in its Python examples using the struct module, which reduces the risk of interpreting binary data as natural language instructions compared to raw text ingestion.

re-drm