The Agent Skills Directory

[COMMAND_EXECUTION]: The skill performs extensive shell operations to manage the development lifecycle.
Evidence: Uses git worktree, git commit, git push, gh pr create, and gh pr merge to handle code changes and repository state.
[REMOTE_CODE_EXECUTION]: The skill dynamically executes shell commands retrieved from the ## Integration Verification section of a plan file (docs/plans/plan-{name}.md).
Evidence: The instruction in Step 4 explicitly states: "Run the smoke commands defined in the plan's ## Integration Verification section."
[DATA_EXFILTRATION]: The skill transmits code and pull request metadata to remote servers.
Evidence: Executes git push origin feature/{plan-name} and gh pr create --body "$(cat .pr-body.md)". These operations target GitHub's official infrastructure, which is standard for the described workflow.
[PROMPT_INJECTION]: The skill ingests untrusted data from external plan files to drive agent behavior and execute commands, creating a vulnerability surface for indirect injection.
Ingestion points: The agent reads docs/plans/plan-{name}.md to extract task descriptions, acceptance criteria, and integration commands.
Boundary markers: None identified. The instructions do not establish clear delimiters or warnings to ignore embedded instructions within the plan content.
Capability inventory: The skill has significant capabilities including shell execution, file system access, and network operations through Git/GitHub.
Sanitization: There is no evidence of sanitization or validation performed on the commands extracted from the Markdown plan before execution.

plan-pipeline-execute