security-hardening

The Security Hardening skill provides an automated defense-in-depth framework designed to protect AI agent workspaces from common vulnerabilities. It empowers agents to act as autonomous security auditors, scanning for configuration risks, accidental secret exposure, and potential prompt injection vectors. This tool is essential for developers, AI engineers, and operators who require a standardized, robust security posture for their agentic workflows without needing deep expertise in cybersecurity.

• Automated Credential Scanning: Detects leaked API keys, tokens, SSH private keys, and hardcoded passwords using regex pattern matching across the entire workspace.

• PII and Data Leakage Prevention: Identifies sensitive personal information such as emails, phone numbers, and physical addresses that are inadvertently stored in public-facing files.

• Config Hardening: Automatically generates and validates security standing orders in AGENTS.md and SOUL.md to enforce behavioral boundaries and external action policies.

• Prompt Injection Defense: Reviews agent instructions for vulnerabilities and suggests defensive hardening measures to prevent malicious external input from manipulating agent behavior.

• File Exposure Review: Checks .gitignore integrity and system configurations to ensure sensitive memory files and environment variables are not exposed to version control or cloud sync services.

• Inputs: The skill processes all local workspace files, specifically targeting configuration files like AGENTS.md, SOUL.md, and source code.

• Outputs: A comprehensive, actionable remediation report detailing severity levels (Critical, High, Warning) and step-by-step resolution instructions.

• Constraints: The tool performs read-only analysis by default. Modification of files requires manual confirmation or specific operator intervention. It does not perform network requests or external data exfiltration.

• Best Practice: Integrate this scan into your periodic heartbeat or cron jobs to ensure continuous compliance and proactive threat detection within your agent environment.