The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes implementation plans and task metadata from the local workspace, which creates a surface for indirect prompt injection.
Ingestion points: Loads plan content from files and state from <plan-path>.tasks.json or .tasks.json.
Boundary markers: The skill relies on natural language instructions for the agent to 'Review critically' and 'STOP executing immediately' if instructions are unclear, rather than technical delimiters.
Capability inventory: The agent has the ability to perform file modifications and execute Git commands via integrated sub-skills.
Sanitization: There is no evidence of automated sanitization or validation of the plan text before the agent attempts to follow its steps.
[COMMAND_EXECUTION]: The skill executes local system commands to manage the development environment.
Evidence: Calls git worktree list to inspect the local Git configuration and determine if a worktree already exists for the plan's branch.

executing-plans