The Agent Skills Directory

[SAFE]: No malicious patterns or security vulnerabilities were identified in the skill documentation or scripts.
[COMMAND_EXECUTION]: The provided Python scripts (risk_assessment.py, compliance_checker.py) are designed to process local data files (CSV) and generate reports in JSON, CSV, or Markdown formats. These scripts use standard Python libraries for data processing and do not perform arbitrary command execution or system modification.
[DATA_EXFILTRATION]: The skill does not perform any network operations. Data processing is confined to the local environment, and there are no instances of sensitive data being sent to external domains.
[PROMPT_INJECTION]: The instructions and trigger phrases are aligned with the stated purpose of the skill (ISO 27001 management). No patterns of safety bypass, role-play jailbreaking, or instruction override were detected.
[INDIRECT_PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface as it ingests untrusted data from local CSV files for report generation. However, the risk is assessed as safe given the following:
Ingestion points: load_controls_from_csv in compliance_checker.py and load_assets_from_csv in risk_assessment.py.
Boundary markers: Data is interpolated into standard Markdown tables without specific boundary delimiters.
Capability inventory: The scripts are limited to file read/write and standard output; they lack network access or dynamic code execution capabilities.
Sanitization: No specific sanitization of CSV input is performed, but the resulting Markdown output is used for reporting purposes within the agent's context and does not trigger dangerous operations.

information-security-manager-iso27001