The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external business documents, such as company-context.md, to provide tailored security advice. This creates a standard surface for indirect prompt injection where instructions embedded in those documents could influence the agent's behavior.
Ingestion points: Processes company-context.md and user-provided queries to generate risk assessments and roadmaps.
Boundary markers: Does not explicitly define delimiters or instructions to the model to ignore potential commands within the context files.
Capability inventory: Uses the agent's capability to execute local scripts (scripts/risk_quantifier.py and scripts/compliance_tracker.py) via the shell.
Sanitization: No specific sanitization or filtering logic is defined for the external context ingestion.
[SAFE]: The skill's Python scripts use only standard libraries for data manipulation and reporting, with no network connectivity or dynamic code execution detected. Metadata and documentation are consistent with the skill's stated purpose, and no obfuscation or privilege escalation patterns are present.

ciso-advisor