The Agent Skills Directory

[SAFE]: No security issues were detected. The skill's operational scope is restricted to information gathering and documentation generation.
[PROMPT_INJECTION]: The skill analyzes untrusted source code and project documentation from external or reference repositories, creating a surface for indirect prompt injection.
Ingestion points: Source code files (referenced via rg/git grep), CONTEXT.md, specs/adr/, and graphify-out/graph.json in the reference and target stacks.
Boundary markers: Absent. There are no explicit instructions to the agent to disregard instructions found within the analyzed source files.
Capability inventory: Reading files via rg and git grep; writing a single markdown artifact to <artifacts-root>/specs/port/. No execution of product code or configuration modification is permitted. The disable-model-invocation: true flag prevents recursive model usage.
Sanitization: None. Content from source files is interpreted directly for mapping purposes without validation or escaping.

port-feature