The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from external websites to generate marketing content. This creates a surface where a malicious website could embed instructions designed to influence the agent's behavior during the video creation or share-copy generation process.
Ingestion points: The references/slim.md file instructs the agent to load websites in a headless browser to extract copy, visuals, and identity elements.
Boundary markers: The instructions do not define explicit delimiters or warnings for the agent to ignore instructions embedded within the scraped website content.
Capability inventory: The skill has the ability to execute shell commands (ffmpeg, hyperframes, node, python3) and write files to the local system.
Sanitization: No explicit sanitization or filtering of the extracted website text is described before it is used to generate the storyboard or share copy.
[COMMAND_EXECUTION]: The skill requires the execution of external CLI tools, which could be exploited if inputs are not properly handled by the agent.
Evidence: The scripts/brag.py script uses subprocess.run to call node and ffprobe. Additionally, references/hyperframes.md provides complex ffmpeg shell commands for the agent to execute.
Analysis: The Python script correctly uses list-based arguments for subprocess.run and avoids shell=True, mitigating direct command injection risks in the script itself. However, the complex shell commands provided in the markdown instructions rely on the agent's ability to safely interpolate file paths and parameters.
[EXTERNAL_DOWNLOADS]: The skill facilitates downloading assets from remote web servers.
Evidence: The instructions in references/slim.md direct the agent to download logos, screenshots, and demo videos from a user-specified website into a local work/ directory.
Analysis: While this is a core feature, it involves fetching untrusted content from the internet. The skill includes a specific warning to "not send project material to an external generation service without authorization," which is a positive security practice.

brag