The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes user-provided files from the exploration/ directory, which presents an indirect prompt injection surface.
Ingestion points: Artifact files (e.g., session briefs, BRDs) located in the exploration/ directory.
Boundary markers: No explicit delimiters are specified for the ingested content in the read operations.
Capability inventory: The skill is permitted to use Bash, Read, and Write tools as defined in the frontmatter.
Sanitization: The skill relies on agent synthesis logic and iterative user review rather than automated sanitization to manage potentially malicious content in source files.

exploration-handoff