The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the AI agent to read and modify local plan artifacts (plan.md) and create review records. These operations are limited to the filesystem and are conducted using the agent's standard file-handling capabilities within the project directory structure (e.g., plans/, specs/, reviews/). No arbitrary shell command execution or system-level modifications are requested.
[DATA_EXFILTRATION]: There are no network operations, such as curl or wget, or any instructions to send data to external servers. The workflow is entirely contained within the local environment.
[EXTERNAL_DOWNLOADS]: The skill does not define any external dependencies or download any scripts or packages at runtime.
[PROMPT_INJECTION]: The instructions are highly prescriptive and define a clear logic for processing inputs (the 'Four Outcomes' framework). While the skill processes user-generated content (plans and specs), it includes specific guidance to treat these as data artifacts (compiler-IR) to be checked against a contract (the spec), which reduces the risk of the agent treating embedded text as new instructions.

review-plan