The Agent Skills Directory

[SAFE]: The skill's primary function is to automate compliance evidence collection using the Bash tool to run industry-standard CLI commands (e.g., aws iam, gcloud projects, az role). These commands are documented as templates for the agent to follow and are restricted to read-only metadata gathering necessary for auditing purposes.
[PROMPT_INJECTION]: The skill processes external data from security scanners and application exports (e.g., npm audit, semgrep, terrascan), which presents a theoretical indirect prompt injection surface. However, this is an inherent and necessary part of its primary auditing purpose.
Ingestion points: references/evidence-collection.md (lines 51, 60, 63
reading output from infrastructure scanners and dependency auditors).
Boundary markers: Absent in templates.
Capability inventory: Bash, Read, and Write tools authorized in SKILL.md frontmatter.
Sanitization: Not explicitly implemented in the provided shell templates.
[COMMAND_EXECUTION]: While the skill contains numerous shell command patterns, they are presented as transparent examples for interacting with legitimate cloud provider APIs and internal services. No suspicious execution patterns like curl | bash or obfuscated payloads were found.

compliance-audit