The Agent Skills Directory

[SAFE]: No malicious patterns, obfuscation, or unauthorized access attempts were identified. The skill operates within its described scope using local files and tools.
[COMMAND_EXECUTION]: The skill executes a local utility script at ${CLAUDE_PLUGIN_ROOT}/scripts/honne to check for the existence of required persona files. This is a standard use of project-specific tooling.
[PROMPT_INJECTION]: The skill ingests user-defined content from local files to guide the AI's persona behavior, creating a surface for indirect prompt injection.
Ingestion points: Step 3 reads content from .honne/personas/antipattern.md, .honne/personas/signature.md, and .honne/personas/judge.md.
Boundary markers: None; the content is mentally applied as system instructions for the duration of the debate rounds.
Capability inventory: The skill can execute the local honne script and read local files.
Sanitization: No sanitization or escaping is performed on the ingested persona content.

crush