The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection by processing instructions from external project files to drive its orchestration logic.
Ingestion points: The skill reads task definitions, wave structures, and requirements from files such as .scratch/<slug>/tasks.md, tech-spec.md, and NOTES.md.
Boundary markers: No explicit delimiters or instructions to ignore embedded commands are provided when the agent interpolates data from these files into its context.
Capability inventory: The skill possesses the capability to modify the codebase (via /file-refactor and /tdd) and trigger deployment actions (via /gitops).
Sanitization: No validation or filtering is applied to the content of the processed files before execution.
Mitigation: The skill enforces a human-in-the-loop requirement for its most sensitive operation, stating that it should only commit when the user explicitly asks and never autonomously.
[COMMAND_EXECUTION]: The skill manages a workflow by invoking several internal slash commands (e.g., /code-graph, /tdd, /prune, /review). These are used for project-specific development tasks and do not involve arbitrary command execution from untrusted sources.

aiops-implement