The Agent Skills Directory

[SAFE]: The skill implements a strictly read-only analysis workflow. It explicitly forbids editing files, switching to execution work, or recommending fixes as primary outputs.
[SAFE]: The instructions emphasize an 'evidence-vs-inference' distinction, requiring the agent to back every claim with concrete repository artifacts (file paths and line numbers).
[SAFE]: The skill uses {{ARGUMENTS}} to ingest user tasks but surrounds this with high-quality constraints that prevent the user input from overriding the read-only contract.
[SAFE]: No external network operations, credential harvesting, or unauthorized privilege escalation patterns were identified in the instructions.

analyze