| name | security-hardening |
| description | Repo-aware security review and remediation workflow for Node.js services using the repository's security memory, commands, and findings process. |
Security Hardening Skill
Use this skill when the task involves:
- reviewing the repository for security issues,
- creating or refining a security hardening plan,
- validating a suspected vulnerability,
- converting a confirmed issue into a tracked finding,
- proposing or implementing a fix for an existing finding,
- re-running review work while avoiding duplicate findings.
This skill is designed to be portable across repositories, while adapting to each repo's local rules, commands, and security folder contents.
Objectives
- Review the codebase in a repo-specific way, not with generic security advice.
- Prioritize practical, high-signal risks first: dependency/supply-chain, exposed services, auth/trust boundaries, secrets/config, command execution, filesystem access, and unsafe network flows.
- Use the repository's own security memory before creating new findings.
- Keep fixes minimal, reviewable, and scoped to one issue at a time.
- Preserve portability: prefer reusable process and evidence over repo-specific assumptions.
Required repo context
Before acting, inspect and use these if present:
.cursor/rules/
.cursor/commands/
.cursor/plans/
security/README.md
security/threat-model.md
security/findings-reviewed.json
security/findings.schema.json
security/accepted-risks.md
security/templates/
- relevant app files such as package manifests, lockfiles, server entrypoints, API routes, process supervision, webhook/email/notification config, and deployment scripts
If some files are missing, continue with the available context and explicitly note what is absent.
Operating principles
- Be repo-specific.
- Do not invent findings without code/config evidence.
- Do not duplicate an existing finding unless the new case is materially different.
- Prefer small, isolated changes over broad refactors.
- Treat security findings as human-reviewed records, not auto-truth.
- When uncertain, propose validation steps instead of overstating risk.
- Respect accepted risks already documented by maintainers.
Priority order
Review in this o