The Agent Skills Directory

[Indirect Prompt Injection] (LOW): The skill is designed to parse repository files and generate a symbol map, which is then used by the agent for reasoning. This creates a surface for indirect prompt injection if the source code contains patterns designed to mislead the agent. Evidence: Ingestion points: Local repository source files processed by ast-grep. Boundary markers: Absent; the JSON schema does not specify delimiters to isolate content from instructions. Capability inventory: The skill relies on external tools (ast-grep, map-validator) but does not contain internal file-write or network operations. Sanitization: None identified in the skill definition.

repo-mapping