The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes Python scripts (check_staleness.py, create_handoff.py) that utilize subprocess.run to interact with Git. These scripts are implemented using list-based arguments, which effectively mitigates shell injection risks. The commands are limited to legitimate project management operations such as checking branch status and commit history.
[DATA_EXPOSURE]: The validate_handoff.py script provides a security-positive feature by scanning generated handoff documents for sensitive patterns such as API keys, private keys, and authentication tokens. This helps ensure that credentials are not inadvertently saved or shared.
[INDIRECT_PROMPT_INJECTION]: The handoff tools process user-generated or agent-generated markdown files (e.g., in .claude/handoffs/).
Ingestion points: Handoff files are read and parsed by check_staleness.py and validate_handoff.py using regex.
Boundary markers: None explicitly implemented; the system relies on standard markdown formatting.
Capability inventory: The scripts have the ability to execute Git commands via subprocess.run based on paths extracted from the files.
Sanitization: The scripts use strict regex matching to extract specific metadata (dates, branches) rather than executing arbitrary text from the documents, which minimizes the risk of indirect injection.

agent-workflow-pro