The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes user-supplied data from files such as research.md and guideline.md to generate social media content. This creates a surface for indirect prompt injection where malicious instructions could be embedded in the research material. Evidence: 1. Ingestion points: External files are read into the agent context (SKILL.md). 2. Boundary markers: No specific delimiters or "ignore embedded instructions" warnings are present. 3. Capability inventory: Access to MCP tools and file system write/copy operations. 4. Sanitization: No content validation or escaping is performed.
[COMMAND_EXECUTION]: The skill performs routine file system management, including creating directories under the outputs/ path and copying existing project files.

write-post