The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the Bash tool to perform comprehensive repository searches using find and grep and to execute local vulnerability scanners like npm audit and pip-audit. These operations are limited to evidence gathering and are consistent with the skill's primary function.
[DATA_EXFILTRATION]: The skill identifies and examines sensitive files such as .env and configuration for secret-management systems (e.g., HashiCorp Vault, AWS KMS) as part of the audit process. The risk of exposure is mitigated by explicit instructions across all modules requiring the agent to redact discovered secret values as [REDACTED] and never include them in the final reports.
[PROMPT_INJECTION]: An indirect prompt injection surface is present because the skill ingests and processes content from arbitrary source files and documentation within the repository. * Ingestion points: File content is read via Read and Bash tools during the audit. * Boundary markers: Ingested content is not wrapped in specific delimiters. * Capability inventory: The skill has access to Bash and Write tools. * Sanitization: While secrets are redacted, the skill does not explicitly sanitize malicious natural language instructions that might be embedded in the audited files. This is a common characteristic of automated analysis tools and is documented here as an inherent risk.

iso27001-audit