The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses uv run python3 to execute multiple internal scripts for process enforcement, such as ds-post-subagent-guard.py, ds-pre-subagent-clear.py, and phase-gate-guard.py. These scripts are located within the vendor's own plugin root and skill directory structure.
[COMMAND_EXECUTION]: Employs dynamic context injection at load time using the !command syntax to execute load-constraints.py. This is used to initialize state-based constraints for the ds-plan workflow.
[COMMAND_EXECUTION]: Utilizes a Monitor command to track the progress of long-running background tasks by tailing local log files.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data (e.g., CSV and Parquet files) during its profiling phase.
Ingestion points: Reads local data files for shape, type, and quality analysis; reads .planning/SPEC.md for task generation.
Boundary markers: Missing; the skill directly processes the contents of these files using standard data science libraries (pandas).
Capability inventory: The skill and its subagents have access to Bash, Read, Glob, Grep, and the ability to run local Python scripts via uv.
Sanitization: No specific sanitization or filtering of data file content is mentioned before it is processed by the agent.
[COMMAND_EXECUTION]: The skill dispatches 'general-purpose' subagents using a Task() function, specifically restricting their capabilities to Read, Glob, Grep, and Bash to prevent unauthorized file modifications during read-only profiling tasks.

ds-plan