The Agent Skills Directory

[SAFE]: The skill's operations are limited to reading local markdown files within the project structure and writing review reports to a specific subdirectory. It performs no network operations and does not execute any external or dynamic code.
[PROMPT_INJECTION]: The skill processes specification documents and decision logs which are external data sources, creating a surface for indirect prompt injection. This risk is assessed as safe because the skill's capabilities are restricted to generating a text-based report and it lacks the ability to execute commands, access secrets, or communicate with the network. Ingestion points: Reads specification artifacts and decision logs as specified in the Workflow section. Boundary markers: The instructions do not specify the use of delimiters or ignore-behavior markers for the ingested content. Capability inventory: The skill can read local files and write markdown files to the repository's reviews folder. Sanitization: No sanitization or validation of the input text is described.

review-spec