The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external draft task prompts to generate refined instructions, creating a surface for embedded instructions to influence the agent's output.- Ingestion Point: User-supplied draft task content provided via the argument-hint.- Boundary Markers: The instructions do not specify the use of delimiters or sanitization protocols to isolate the input content from the skill's logic.- Capability Inventory: The tasks generated by this skill include capabilities for file modification, git repository operations, and execution of local verification scripts.- Sanitization: The skill incorporates the CLEAR (Concise, Logical, Explicit, Adaptive, Reflective) and CoVe (Chain of Verification) frameworks, which act as qualitative filters and verification gates for the generated output.- [COMMAND_EXECUTION]: The skill instructs agents to include and execute system commands as part of task completion and verification workflows.- Evidence: The task template explicitly includes git add and git commit commands for managing outputs.- Evidence: Migration tasks are required to execute a specific local script using uv run plugins/development-harness/scripts/verify_migration_fidelity.py to ensure data integrity.

clear-cove-task-design