The Agent Skills Directory

[SAFE]: The skill serves as a process framework for manual and agent-assisted repository maintenance. It does not perform external network calls or require access to sensitive user data.
[COMMAND_EXECUTION]: The workflow relies on standard git commands (git log, git diff) to identify intentional changes in a fork. These are executed locally and are appropriate for the task.
[INDIRECT_PROMPT_INJECTION]: The skill ingests micropatch.md files which act as implementation guides. While these files represent an external source of instructions, the risk is managed by the skill's emphasis on intent-based reasoning and the 'Definition of Done' verification process.
Ingestion points: The micropatch.md file content is read during the apply phase.
Boundary markers: Absent.
Capability inventory: File system writes and local execution of test commands.
Sanitization: Absent. This is a low-risk architectural surface typical of code generation and maintenance tools.

micropatch