The Agent Skills Directory

[SAFE]: The skill operates under a strict safety contract, prioritizing read-only inspections and explicitly prohibiting dangerous activities such as brute force or denial-of-service tests without specific authorization.
[INDIRECT_PROMPT_INJECTION]: The skill inherently processes untrusted code and configuration files during audits, which presents a surface for indirect prompt injection, but this is thoroughly mitigated by the skill's operational design.
Ingestion points: The skill ingests backend source code, route definitions, middleware configurations, and package manifests (SKILL.md, Establish scope section).
Boundary markers: The skill instructions include a mandatory 'Safety contract' and require a concise proposal with user confirmation before any state-changing actions are performed (SKILL.md).
Capability inventory: The skill has the ability to modify files, install packages, and change configurations, but only after explicit user selection and approval (SKILL.md, Implement selected work only section).
Sanitization: Not applicable as the skill performs analysis and recommends fixes based on identified vulnerabilities rather than executing the analyzed code directly.

nodejs-security