The Agent Skills Directory

Indirect Prompt Injection Surface: The skill is designed to process external, untrusted data such as interview notes and email exports. This creates a potential surface where malicious instructions embedded within that data could attempt to influence the agent's behavior. This is a common consideration for skills that perform data analysis.
Ingestion points: The skill receives data through user input (matter name or slug, plus pasted or attached data) as defined in SKILL.md.
Boundary markers: No specific delimiters or "ignore" instructions for the processed data are defined in this wrapper script.
Capability inventory: The skill processes data against criteria and logs results, with core logic delegated to the internal-investigation reference skill.
Sanitization: No explicit sanitization or filtering of the input data is mentioned in the instruction set.
Local Configuration Loading: The skill instructions involve loading a configuration file (CLAUDE.md) from a specific subdirectory within the user's home folder (~/.claude/plugins/config/). This is a standard mechanism for plugins to access their internal rules and operational settings.

investigation-add