The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill analyzes repository code and git differences, which represent an ingestion surface for potential indirect instructions embedded in code comments or data.
Ingestion points: Git status, git diff, and the content of files modified during the session (SKILL.md Gate 0).
Boundary markers: The skill requires an explicit user confirmation step (Gate 0-5) before staging or committing any changes, and it uses session history to define an authoritative scope.
Capability inventory: The agent performs repository operations including git add, git commit, git push, and branch cleanup. It also executes local project commands for linting, testing, and building.
Sanitization: The workflow requires the agent to build an explicit list of 'in-scope' files and protects 'out-of-scope' files from modification or accidental staging.
[COMMAND_EXECUTION]: The skill relies on standard shell commands for version control and development lifecycle management.
Evidence: Instructions include the use of git status, git branch, git log, git diff, and git add for specific files. It also triggers project-defined scripts for verification (tests, lint, build) as part of the production-readiness check.

finalize-and-commit