The Agent Skills Directory

[COMMAND_EXECUTION]: The SKILL.md file contains a 'Quick Start' section that instructs the execution of Python scripts located at 'scripts/risk_quantifier.py' and 'scripts/compliance_tracker.py'. As these files are missing from the provided skill payload, the logic being executed is unverified.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection through its integration with external data sources. * Ingestion points: The agent is directed to read 'company-context.md' and various files in the 'references/' directory. * Boundary markers: There are no instructions providing delimiters or warnings to ignore potentially malicious embedded instructions within these data sources. * Capability inventory: The agent has the ability to execute shell commands via the referenced Python scripts and can invoke other agent roles using the [INVOKE] syntax. * Sanitization: The instructions do not specify any validation or sanitization of the content retrieved from the ingested files.
[DATA_EXFILTRATION]: While not explicitly exfiltrating data, the skill instructions require the agent to process 'crown jewel' data and regulatory notification timelines from 'company-context.md'. This established ingestion pattern, combined with the command execution capabilities, creates a significant risk surface for sensitive data exposure.

ciso-advisor