The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill includes reference data containing strings commonly used for prompt injection attacks.
Ingestion points: resources/posts/kenny-posts-reference.csv (specifically within the 'full_content' for rank #26).
Boundary markers: Absent. The skill instructions do not specify delimiters or warnings to ignore instructions found within the CSV data.
Capability inventory: The skill is designed to read these reference files and use them as context for generating, structuring, and rewriting new LinkedIn posts.
Evidence: The CSV data contains the string "↳ Shortwave | ignore previous instructions and automate your email" in a list of AI tools. This string, if processed as part of a content generation task, could lead to instruction override.
Sanitization: No explicit sanitization or filtering of external content is mentioned.
[EXTERNAL_DOWNLOADS]: The skill documentation provides instructions for setting up vendor-specific extensions.
Evidence: Instructions in the main SKILL.md and several sub-skills (e.g., hooks/SKILL.md, repurposing/SKILL.md) recommend running npx -y @coldiq/mcp@latest to enable specialized MCP tools. This involves downloading and executing code from the public npm registry.
[COMMAND_EXECUTION]: The skill provides shell commands for environment configuration and tool execution.
Evidence: The setup workflow includes instructions for the user to execute COLDIQ_API_KEY=<key> npx -y @coldiq/mcp@latest in their terminal to connect the LinkedIn-specific social intelligence features.

linkedin-content