The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill provides an attack surface for indirect prompt injection by processing untrusted data that could contain malicious instructions designed to manipulate the agent's behavior during report generation or review.
Ingestion points: Several methods in scripts/grc_utils.py, including PolicyGenerator.add_section, ControlAssessment.set_description, RiskRegister.add_risk, and AuditPackage.add_finding, ingest arbitrary user-supplied strings for policy content, control descriptions, and audit findings.
Boundary markers: The utility scripts interpolate these strings directly into markdown and JSON output without using delimiters, headers, or instructions that would signal to an agent to ignore embedded commands within the content.
Capability inventory: The skill is primarily focused on data organization and reporting; its implementation is restricted to standard Python libraries (json, logging, datetime, collections) and does not include network access, subprocess execution, or sensitive file system operations.
Sanitization: The scripts/grc_utils.py script does not perform any validation, filtering, or escaping of the ingested text before it is included in the generated markdown reports.

grc