The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes standard version control commands using git and the GitHub CLI (gh). These operations include checking branch status, staging files, committing, pushing to remote repositories, and creating Pull Requests. All such operations are strictly gated by human confirmation requirements.
[INDIRECT_PROMPT_INJECTION]: The skill ingests external data from the file system (.ways/state.json) and git metadata (branch names). While this data could potentially contain malicious instructions, the skill mitigates this risk by requiring the user to manually review and confirm the status, diffs, and commit messages (Steps 2, 3, and 4) before any action is performed.
Ingestion points: .ways/state.json, git branch output, and git diff output.
Boundary markers: The instructions explicitly demand that the agent pauses and waits for user confirmation before executing git add, git commit, git push, or gh pr create.
Capability inventory: File system reads, git CLI, and gh CLI.
Sanitization: Relies on mandatory human review of the generated content and proposed commands before execution.

fmway-pr