The Agent Skills Directory

[SAFE]: The skill operates as a purely instructional and interactive utility, enforcing human-in-the-loop oversight for every phase of the plan review through the AskUserQuestion tool.
[COMMAND_EXECUTION]: The instructions suggest analyzing git log to understand historical context for code changes, which is a standard development practice.
[INDIRECT_PROMPT_INJECTION]: The skill processes project plans and source code which are external, untrusted inputs.
Ingestion points: Reads files and plan documentation using Read, Grep, and Glob tools.
Boundary markers: None explicitly defined in the instructions for separating input data from agent context.
Capability inventory: Access is strictly limited to read-only file system operations and interactive communication.
Sanitization: The skill lacks access to sensitive capabilities such as network exfiltration, arbitrary code execution, or file writing tools that could be leveraged by malicious data.

plan-gate