The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill contains several tools designed to process external, potentially untrusted data. For instance, gdpr_compliance_checker.py scans project directories, and risk_assessment.py ingests asset lists from CSV files. If these files contain malicious instructions disguised as data, the agent could potentially execute them during its analysis of the tool's output.
Ingestion points: File system paths (directories) in gdpr_compliance_checker.py, asset inventories (CSV) in risk_assessment.py, and process definitions (JSON) in audit_schedule_optimizer.py and capa_tracker.py.
Boundary markers: The instructions lack explicit delimitation or warnings to ignore instructions embedded within the data being processed.
Capability inventory: The skill utilizes subprocess calls to execute local Python scripts across multiple modules (e.g., regulatory_tracker.py, isms_audit_scheduler.py, risk_matrix_calculator.py) which perform file-system operations and data processing.
Sanitization: There is no evidence of sanitization or filtering of the content being analyzed before it is returned to the agent's context.
[COMMAND_EXECUTION]: The skill documentation frequently prompts the user and the agent to execute numerous local Python scripts (e.g., python3 scripts/risk_matrix_calculator.py, python3 scripts/isms_audit_scheduler.py, python3 scripts/capa_tracker.py). While the skill documentation asserts these tools only use the Python standard library, the execution of local scripts by the agent represents a direct utilization of the command execution capability available to the environment.

ra-qm-skills