The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides Python code snippets for data collection and instructs the agent to execute them using libraries like TikTokApi and instaloader (Step 4). It also utilizes command-line tools like yt-dlp for metadata extraction.
[EXTERNAL_DOWNLOADS]: The instructions require installing several third-party libraries from public registries, including TikTokApi, instaloader, and playwright, which are necessary for the scraping functionality.
[DATA_EXFILTRATION]: The skill's bootstrap process involves a broad search and read of local files, including CLAUDE.md, MEMORY.md, and the broader codebase and analytics events, to automatically discover product details and context (Step 2).
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests and processes untrusted external data from social media comments and video descriptions to perform its analysis (Step 4 and 5).
Ingestion points: Social media video descriptions and user comments (SKILL.md Step 4).
Boundary markers: None identified; the skill directly analyzes the ingested text strings.
Capability inventory: Execution of Python scripts, file writing for state management, and network operations for scraping.
Sanitization: No sanitization or filtering of the external social media content is documented before processing.

trendwatch