The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes a bash snippet to verify that the required TIKHUB_API_KEY environment variable is defined before execution.
[COMMAND_EXECUTION]: It utilizes vendor-specific CLI tools like tikhub-find-endpoint to dynamically locate API paths, which is expected behavior for an agent skill integrated with a specific platform.
[PROMPT_INJECTION]: The skill is subject to an indirect prompt injection surface because it processes untrusted content from social media platforms (TikTok, Instagram, Twitter, etc.).
Ingestion points: Workflow step 3 in SKILL.md identifies that posts and comments are collected into a list for analysis.
Boundary markers: No specific delimiters or "ignore instructions" warnings are defined for the ingested social media text.
Capability inventory: The skill is designed for sentiment analysis, theme clustering, and report generation; it does not possess dangerous capabilities like arbitrary file writing or shell execution using the ingested data.
Sanitization: The skill performs relevance filtering but does not explicitly mention sanitizing ingested content for malicious instructions.
[SAFE]: The skill uses legitimate vendor infrastructure (TikHub) and standard API patterns for its primary functionality.

social-listening