The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes repository content, including code, tool outputs, and existing documentation, to generate new material. This represents a potential surface for indirect prompt injection from untrusted files within a codebase. However, the instructions emphasize faithful mapping, quality checks, and validation of paths, which mitigates accidental compliance with embedded instructions.
Ingestion points: Processes file paths, repository commands, and actual code/tool behavior within the workspace.
Boundary markers: Instructions require the agent to distinguish evidence-backed causes from symptoms and verify instructions against current repository state.
Capability inventory: The skill requests workspace-write access to implement documentation changes.
Sanitization: The skill lacks explicit sanitization instructions for external content but focuses on technical accuracy and safety notes for error-prone steps.

documentation-engineer