The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses local shell commands such as git diff and find to identify files changed in a specific commit or located within a directory. These operations are strictly scoped to the user's project and include explicit exclusions for sensitive directories like .git, .env, .venv, node_modules, and build artifacts.
[REMOTE_CODE_EXECUTION]: For large sets of files, the skill delegates work to sub-agents using a task-based parallelization pattern. These sub-tasks are restricted to specific file blocks and follow the same transformation rules as the main agent, ensuring consistent and controlled processing.
[DATA_EXPOSURE]: The skill implements a safety-first approach where it categorizes AI-detected artifacts into tiers. High-confidence artifacts are stripped, while behavior-changing items (such as potentially hallucinated package references or logic placeholders) are flagged for human review to prevent accidental code corruption or data loss.
[SAFE]: All external references, including the license and documentation links, point to legitimate domains associated with the skill's author and official project repositories. The skill does not perform any unauthorized network communications or credential harvesting.

text-human