The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes task descriptions and requirements from local project files which could be influenced by external contributors (e.g., via malicious pull requests).
Ingestion points: The skill reads conductor/tracks/{trackId}/plan.md and conductor/tracks/{trackId}/spec.md to identify tasks and requirements.
Boundary markers: Absent. The agent is instructed to parse these files directly for task descriptions and phase structures.
Capability inventory: The skill can execute shell commands (git commit, npm test, pytest) and modify local files.
Sanitization: Absent. Content parsed from these files is used directly in prompts and shell commands.
[COMMAND_EXECUTION]: The skill constructs git commit commands by interpolating task descriptions parsed from the implementation plan.
Evidence: The instruction git commit -m "{commit_prefix}: {task description} ({trackId})" in SKILL.md directly uses string interpolation. If a task description in plan.md contains shell metacharacters (e.g., ;, &&, `), it could lead to command injection or shell behavior manipulation during the commit process.

conductor-implement