The Agent Skills Directory

[DATA_EXFILTRATION]: The skill is designed to discover and inspect sensitive files and hardcoded credentials during the audit process.
Sensitive file access: The instructions guide the agent to read and grep through sensitive locations including .env files, SSH private keys (.pem, .key), and credential files (credentials.json).
Credential scanning: It explicitly searches for high-value secrets such as API keys, cloud provider tokens (e.g., AKIA), and GitHub tokens.
[PROMPT_INJECTION]: The skill exhibits an attack surface for Indirect Prompt Injection by processing untrusted data from external sources.
Ingestion points: The skill fetches and processes content from the public web via firecrawl_search and firecrawl_scrape, as well as production logs from Sentry.
Boundary markers: No specific delimiters or instructions are provided to the agent to treat external content as untrusted or to ignore embedded instructions.
Capability inventory: The agent possesses significant capabilities including codebase read access and MCP tool execution which could be abused if the agent follows instructions hidden in scraped content.
Sanitization: External content is not sanitized or filtered before being incorporated into the agent's context.
[COMMAND_EXECUTION]: The skill instructs the agent to execute various third-party command-line utilities for dependency auditing.
CLI Tools: Recommends running npm audit, pip-audit, cargo audit, govulncheck, and bundle audit on the local system.
Context: While these are standard security tools, they involve executing shell commands on the user's environment.

audit-security