The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides the agent with extensive instructions to execute shell commands using the Bash tool for security reconnaissance (e.g., nmap, dig, subfinder), exploitation testing (e.g., sqlmap), and infrastructure auditing (e.g., aws-cli, gcloud, kubectl). These commands are integral to the skill's purpose as a security reviewer but represent a significant capability surface.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted content such as source code, logs, and configuration files. This creates a surface for indirect prompt injection attacks where malicious instructions embedded in the analyzed data could attempt to influence the agent's behavior.
Ingestion points: The skill performs file reading and analysis using Read, Grep, and Glob tools throughout the 'Scan' and 'Review' phases (defined in SKILL.md).
Boundary markers: The instructions do not define specific delimiters or "ignore instructions" wrappers for the content retrieved from external files during analysis.
Capability inventory: The agent possesses powerful capabilities including Bash (shell execution), Read (file system access), and the ability to interact with cloud providers and network services.
Sanitization: No explicit logic for sanitizing or escaping the content of reviewed files is provided in the instructions.
[EXTERNAL_DOWNLOADS]: The reference files (sast-tools.md, secret-scanning.md, infrastructure-security.md) provide instructions for installing numerous third-party security tools from standard package registries (NPM, PyPI, Go) and official GitHub repositories. These include well-known utilities such as Semgrep, Bandit, Trivy, Gitleaks, and Checkov. These downloads target official sources and are documented neutrally as part of the tool setup guidance.

security-reviewer