Security Hardening Engineering Skill
Gives coding agents a structured defense workflow for untrusted input, authentication, sensitive data, and external service integrations.
The skill explicitly requires least privilege, human approval, input validation, secret protection, dependency auditing, and constrained agent/tool permissions. It lacks a general rollback plan, execution scoping, data-flow disclosure, and a consistent confirmation protocol for high-risk actions; it also depends on an unprovided reference file, so points are deducted.
The workflow, tiered rules, examples, and verification checklist are mostly consistent, and the SSRF example explicitly acknowledges a TOCTOU limitation. However, there are no committed tests for the key paths, standardized abnormal-input feedback, or reproducible execution evidence; the score remains within the static-review ceiling.
Triggers clearly cover user input, authentication, sensitive data, and external integrations, making the skill relevant to common web-security work. Its scope is broad, non-fit cases and input/output contracts are not sufficiently defined, semantic exclusions are limited, and Chinese-language or mainland-China reachability is not addressed, so points are deducted.
The material is well layered with overview, triggers, process, examples, anti-rationalization guidance, red flags, and verification; the README and MIT license add installation and maintenance context. However, skill versioning, changelog, concrete update path, and the contents of the referenced security-checklist.md are not supplied, leaving hidden assumptions.
It offers directly reusable patterns for authentication, injection, XSS, SSRF, uploads, rate limiting, secrets, and LLM security, giving it practical core value. It remains generic guidance rather than a repository-specific executable result, and example coverage and adaptation require human review; static evidence does not justify a higher score.
Repository CI validates skill structure, routing evaluations, and installation, but the supplied tests do not establish the correctness of this skill's key security paths or examples. There are limited auditable procedures and citation cues, but no skill-specific independent reproduction or corroborating evidence, so the score is low.
- The skill contains high-impact security guidance; examples and checklists should be reviewed by qualified practitioners against the actual framework, deployment environment, and threat model.
- The referenced references/security-checklist.md was not provided, so its content, availability, and consistency cannot be assumed.
- The SSRF example acknowledges DNS-rebinding TOCTOU risk; an allowlist and DNS check alone do not cover every deployment scenario.
- The CI evidence primarily validates repository structure and generic evaluations; it does not demonstrate that this security skill passed practical security testing.
What does this skill do, and when should you use it?
This is the standalone security-and-hardening skill in addyosmani/agent-skills, focused on web application engineering. It starts with threat modeling and applies a three-tier boundary system to input validation, authentication, authorization, secrets, dependencies, and AI/LLM features. It includes OWASP prevention patterns, SSRF defenses, upload safety, rate limiting, supply-chain hygiene, and post-implementation verification. It fits teams that want coding agents to apply repeatable security checks, but the supplied material presents it as an instructional Markdown workflow rather than an automated security scanner.
It guides an agent to map trust boundaries and assets, apply STRIDE, write abuse cases, and review input validation, parameterized queries, output encoding, HTTPS, security headers, cookies, password hashing, authorization, CORS, rate limits, secrets, and error responses. It supplies TypeScript patterns for Zod boundary validation, upload restrictions, SSRF URL allowlists, dependency-audit triage, install-script controls, and defensive LLM-output handling. After security-sensitive implementation, it requires checks for audit findings, secrets, permissions, headers, error disclosure, rate limiting, SSRF, and model-output handling.
- A Web API engineer can use it when accepting form data, request bodies, uploads, or webhooks to design boundary validation and abuse tests.
- A team implementing login, sessions, password resets, or permissions can use it to review authentication, authorization, cookies, and rate limits.
- A developer handling payment data, PII, passwords, or API keys can use it to plan protection for storage, transport, logs, and response fields.
- A team adding third-party integrations, URL imports, or link previews can use it to assess SSRF, CORS, timeouts, and allowlists.
- An engineer building chatbots, RAG, or agent features can use it to constrain prompt injection, model output, tool permissions, data isolation, and resource consumption.
What are this skill's strengths and limitations?
- Covers a complete workflow from threat modeling through post-implementation verification.
- Provides concrete patterns for injection, XSS, authentication, access control, SSRF, dependency supply chains, and LLM security.
- Clearly separates mandatory actions, human-approval items, and prohibited practices.
- Includes practical TypeScript examples and a security review checklist.
- The supplied material shows no automated scanner, test suite, or execution results, so it should not be treated as a vulnerability-detection tool.
- Some detail is delegated to the separately referenced references/security-checklist.md, whose full contents are not supplied here.
- Examples span multiple ecosystem components without prescribing one technology stack, so implementation still requires project-specific validation.
- It provides guidance and review requirements, with no evidence that it automatically handles approvals, secret rotation, or dependency upgrades.
How do you install this skill?
This skill is part of a repository containing 24 skills. Install the individual skill with the Skills CLI:
npx skills add addyosmani/agent-skills --skill security-and-hardening
The README also documents collection installation for Claude Code, Codex, Cursor, Antigravity CLI, Gemini CLI, Windsurf, OpenCode, GitHub Copilot, and Kiro; the supplied material does not provide a complete per-platform command for installing only this skill. The repository uses the MIT license.
How do you use this skill?
Trigger it in an Agent Skills-compatible coding agent when handling user input, authentication, sensitive data, uploads, webhooks, payment data, or external integrations. For example:
“Threat-model this webhook API first, then review input validation, signature verification, SSRF, rate limiting, secrets, and error responses.”
Use the SKILL.md sections for Always Do, Ask First, Never Do, verification, and red flags as the review structure.
How does this skill compare with similar options?
The README explicitly names Superpowers and Matt Pocock’s skills as comparison points and links to docs/comparison.md; the supplied material gives no specific comparison conclusion for this individual security-and-hardening skill.