The Agent Skills Directory

[DYNAMIC_EXECUTION]: The script scripts/run.py calculates the repository root at runtime by searching upward for an AGENTS.md file. It then dynamically modifies sys.path to import the tooling.common module. This represents dynamic loading from a computed path.
[INDIRECT_PROMPT_INJECTION]: The skill provides a filesystem modification surface that could be exploited if the agent is manipulated into providing an unintended target path.
Ingestion points: The --workspace command-line argument in scripts/run.py, which is derived from agent context and instructions.
Boundary markers: SKILL.md instructs the agent to target workspaces/<name>/ and explicitly warns against pointing the script at the repository root.
Capability inventory: The script uses copy_tree to write multiple template files and directories (such as STATUS.md, UNITS.csv, and various subdirectories) to the target location.
Sanitization: While the script uses Path.resolve() to normalize the target path, it does not implement a check to ensure the resolved path remains within the designated workspaces/ directory, relying instead on the agent's adherence to instructions.

workspace-init