The Agent Skills Directory

[SAFE]: No security vulnerabilities were identified. The skill implements a robust 'Self-Correction Protocol' in every mode to prevent the agent from executing implementation code or modifying the codebase. Filesystem operations are restricted to discovery (ls, Glob, Grep) and reading requirements, while write operations are targeted at a dedicated output directory (.mula/outputs/). Interactions with Jira are performed via the torque-tool CLI, with built-in safeguards requiring user confirmation before any issue deletion or modification. The skill's architecture naturally mitigates risks by separating planning from execution.

torque-refine