The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external TikTok profiles and video captions, which could contain malicious instructions designed to influence the agent's output.
Ingestion points: External data is ingested from public TikTok profile pages and video captions as specified in references/workflow.md.
Boundary markers: The SKILL.md file defines explicit "Output boundary" rules to separate raw observations from the agent's calculations and judgments.
Capability inventory: The skill possesses capabilities to read files, search for creators, and extract video captions via the kolsprite MCP tool defined in agents/openai.yaml.
Sanitization: The references/workflow.md file defines a normalization and validation process for TikTok URLs to prevent arbitrary ingestion.
[SAFE]: The skill connects to an external MCP endpoint at https://mcp.kolsprite.com/mcp. This is a vendor-owned resource (dev-kolsprite) required for the skill's data enrichment features and does not present an unusual security risk.

tiktok-account-audit