Works with the AI Agents you already use
Claude Code
Cursor
Codex CLI
Muse
Dots+22 moreClaude CodeCursorCodex CLIMuseDots
OpenClaw
Manus
Grok
Grok Bot+18 agentsSecurity Skills
Composable security skills for threat modeling, security reviews, secrets hygiene, dependency audits, access control, web hardening, and incident response.
Free
security-skills
Example session with this skill installed
/auth-access-review Check our Next.js and Supabase API routes in app/api/ for broken object-level authorization (BOLA), missing session checks, and tenant isolation gaps.
- Read your context and instructions
- Compiled the security-
Security Review Summary: Authentication & Access Control
Inspected 8 route handlers under app/api/ and database policies in supabase/migrations/. Identified
1 Critical and
1 High severity access control vulnerability with confirmed reachability.
| Severity | Finding | Location | Status |
|---|---|---|---|
| Critical | Broken Object-Level Authorization (BOLA) | app/api/projects/[id]/route.ts:28 | Confirmed Exploitable |
| High | Service Role Key Used in Client-Facing Route | app/api/exports/route.ts:14 | Confirmed Exploitable |
| Medium | Missing Rate Limiting on Password Reset | app/api/auth/reset/route.ts:9 | Needs Verification |
1. [CRITICAL] Broken Object-Level Authorization (BOLA) on Project Deletion
- File:
app/api/projects/[id]/route.ts(Lines 22–35)
Issue: The DELETE handler verifies that a user session exists via supabase.auth.getUser(), but deletes the record matching params.id without verifying owner_id or organization membership.
- Remediation: Scope the query to the authenticated user's ID and enforce Row Level Security (RLS):
const { data: { user } } = await supabase.auth.getUser();
if (!user) return NextResponse.json({ error: 'Unauthorized' }, { status: 401 });
const { error } = await supabase
.from('projects')
.delete()
.eq('id', params.id)
.eq('owner_id', user.id); // Enforce ownership check
Connects securely to your tools. The creator never sees your data.
About this skill
Why Security Skills?
Security guidance is most useful when it is available at the exact moment a developer needs it. Security Skills packages 12 focused, practical security workflows for AI coding agents that can be run during everyday engineering work.
Instead of generic advice, these skills instruct your coding agent to inspect actual project files, confirm vulnerability reachability, separate facts from assumptions, and produce actionable remediation plans.
Included Workflows (12 Composable Skills)
🛠️ Setup & Context
/setup-security-skills: Detects your project stack and available scanners, gathers missing context, and generates a tailoredSECURITY-CONTEXT.md.
📐 Design & Code Review
/security-grill: Interviews you about assets, entry points, trust boundaries, identity, and failure modes before writing code./threat-model: Generates a focused threat model using assets, actors, trust boundaries, and the STRIDE framework./security-review: Reviews diffs, branches, pull requests, or files for exploitable vulnerabilities (SQLi, XSS, SSRF, path traversal, etc.)./auth-access-review: Audits authentication, session handling, BOLA/IDOR, object ownership, roles, and multi-tenant isolation.
🛡️ Infrastructure, Web & Pipelines
/web-hardening: Reviews security headers, CORS, CSRF, cookies, CSP, file uploads, and rate limiting (includes dedicated references for Django, Next.js/Supabase, and Express/Fastify)./api-security: Audits REST, GraphQL (introspection, depth limits), and gRPC APIs for input validation, rate limiting, and excessive data exposure./container-security: Checks Dockerfiles, runtime privileges, resource limits, and Kubernetes manifests (RBAC, network policies, pod security)./cicd-security: Hardens GitHub Actions and CI/CD pipelines against secrets exposure, overly permissive tokens, and supply-chain risks.
🚨 Prevention & Incident Response
/secrets-hygiene: Finds, rotates, prevents, and cleans up exposed credentials in the working tree, logs, and Git history./dependency-audit: Audits lockfiles and dependencies for known CVEs, typosquatting, and malicious install scripts./incident-response: Structured workflow to contain, preserve evidence, scope, eradicate, and recover from a suspected compromise.
What's included (12 skills in this plugin)
api-security
Review REST, GraphQL, and gRPC APIs for security issues including broken authentication, excessive data exposure, lack of rate limiting, mass assignment, injection through query parameters or request bodies, improper input validation, and insecure API versioning. Use when the user mentions API security, REST hardening, GraphQL introspection, API gateway, API keys, OAuth scopes, or asks whether an API endpoint is safe. Do not use for web-page hardening without an API concern; use `/web-hardening`.
auth-access-review
Review authentication, authorization, and multi-tenant isolation for broken access control such as IDOR, missing permission checks, privilege escalation, mass assignment, and cross-tenant data leaks. Use when the user mentions roles, permissions, login, JWT, sessions, tenant isolation, admin panels, object ownership, or asks whether user A can see or change user B's data. Use when a route, endpoint, view, background job, export, or policy changes authorization. Do not use for a general security review when no access-control behavior is in scope.
cicd-security
Review CI/CD pipelines for security issues including secrets exposure in logs or environment, untrusted code execution in pull requests, missing artifact integrity checks, over-permissive workflow permissions, and supply-chain attacks through actions or plugins. Use when the user mentions CI/CD security, GitHub Actions hardening, pipeline secrets, build security, artifact signing, or asks whether their deployment pipeline is safe. Do not use for application-code vulnerabilities; use `/security-review`.
container-security
Review Docker, OCI, and Kubernetes configurations for security issues including privileged containers, root users, exposed secrets, unscanned images, missing network policies, and insecure defaults. Use when the user mentions Docker security, Dockerfile review, container hardening, Kubernetes RBAC, pod security, image scanning, or asks whether a container deployment is safe. Do not use for application-code vulnerabilities inside the container; use `/security-review`.
dependency-audit
Audit project dependencies for known vulnerabilities, abandoned packages, malicious or typosquatted packages, and supply-chain risk, then propose safe upgrades. Use when the user asks about vulnerable packages, npm audit, pip-audit, Dependabot alerts, outdated libraries, lockfiles, release readiness, or deployment risk. Do not use for application-code vulnerabilities unrelated to dependencies; use `/security-review`.
incident-response
Guide the first hour of an active or suspected security incident such as a compromised account, breached host, suspicious access, defaced site, malware, or exposed credential with possible misuse. Contain, preserve evidence, scope impact, eradicate persistence, recover, notify the right people, and write the follow-up. Use when the user says they were hacked, breached, taken over, attacked, or sees suspicious activity. Use this before `/secrets-hygiene` when a leaked secret may already have been used.
secrets-hygiene
Find, prevent, and clean up leaked secrets such as API keys, tokens, passwords, private keys, service-account files, and `.env` files in the working tree or Git history. Use when the user mentions a secret, credential, API key, token, `.env`, leaked key, exposed password, public repository, or committing configuration. Use before pushing or deploying when credential exposure is possible. Do not use as the primary flow for an active compromise; use `/incident-response` first and then hand off here for secret rotation and cleanup.
security-grill
Interview the user about the security of a plan, feature, or design until the important attack paths, assets, identities, trust boundaries, and recovery decisions are resolved. Use when the user wants to stress-test a design, says "grill me on security", or is about to build something handling authentication, payments, uploads, secrets, or personal data. Do not use for a finished code-diff audit; use `/security-review`.
security-review
Review a diff, branch, pull request, or set of files for exploitable security vulnerabilities. Check injection, broken authentication or authorization, data exposure, unsafe deserialization, SSRF, file handling, secrets, and misconfiguration. Use when the user asks for a security review, audit, vulnerability check, exploitability assessment, "is this safe", or before merging code that changes authentication, user input, file handling, database queries, or external requests. Do not use for general code quality or performance reviews without a security concern.
setup-security-skills
Configure the security skills for a repository by recording its stack, authentication, hosting, report location, severity scale, incident owner, and available scanners in SECURITY-CONTEXT.md. Use once before the other security skills, or when the user asks to configure, set up, or reconfigure security skills. Do not use for a one-off security review without repository configuration.
threat-model
Produce a lightweight, written threat model for a system or feature using assets, actors, trust boundaries, attack paths, and STRIDE. Use during design, when the user asks for a threat model, security design review, trust-boundary analysis, or attack-surface analysis, or after a `/security-grill` session. Do not use for a code-diff vulnerability audit; use `/security-review`.
web-hardening
Harden a web application's configuration, covering HTTPS, security headers, CORS, CSRF, cookies, CSP, rate limiting, file uploads, error handling, and production settings. Use when the user is preparing to deploy, asks about headers, CORS errors, cookies, CSRF, CSP, uploads, rate limits, "is my site secure", or works in Django/DRF, Next.js, React, or Supabase. Do not use for object-level authorization or tenant-isolation reviews; use `/auth-access-review`.
How to use
Skills on Agensi run inside your AI agent. Copy the prompt, paste it into your agent, and it connects to Agensi and installs this skill for you. If a payment is needed, your agent shows it to you first.
Reviews
1 people have installed this skill.
Trust & safety
Security scanned
Verified clean 2 days ago
- Free to download with an account
Requires
git
Node.js