social-media-kit
Pass
Audited by Gen Agent Trust Hub on Oct 24, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The script scripts/cardrender.py utilizes subprocess.Popen to execute the ffmpeg command for video encoding. The implementation is safe as it uses a fixed list of arguments rather than a shell string and pipes raw video data via stdin, which is a standard and secure practice for media rendering.
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied social media post content to generate media assets and automate publishing across platforms like X, LinkedIn, and Reddit. This creates a potential surface for indirect prompt injection if the input content contains instructions designed to manipulate the agent's browser automation steps.
- Ingestion points: Social media post content and environment variables such as BRAND_URL and QUOTE_ATTRIBUTION are processed in SKILL.md and referenced scripts (e.g., linearcard.py).
- Boundary markers: No explicit delimiters or instructions are defined to separate post content from the agent's instructions, though standard LLM guardrails apply.
- Capability inventory: The skill leverages browser automation via Chrome DevTools Protocol (CDP), local file system writes for asset generation, and shell command execution for video processing.
- Sanitization: No specific sanitization of input text was found before its interpolation into cards or use in browser automation scripts.
Audit Metadata