The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external social media profiles and search engine results, creating an attack surface for indirect instructions.
Ingestion points: The skill scrapes competitor profiles using Firecrawl and Playwright, and searches for trends via SerpApi (SKILL.md).
Boundary markers: There are no explicit delimiters or instructions provided to the agent to distinguish between research data and system instructions.
Capability inventory: The skill performs file-write operations to save context and generated captions in the context/ and outputs/ directories (SKILL.md).
Sanitization: No sanitization, filtering, or validation of the content retrieved from external sources is described.
[SAFE]: No malicious patterns were identified. The skill's behavior, including tool usage and file access, aligns with its primary purpose as a social media content generator. The use of external tools for competitor and trend research is a legitimate function of the skill and is implemented through standard tool invocations.

caption-writer