a2a-security-review
Review agent-to-agent (A2A) / multi-agent-system trust: peer identity and authentication, message integrity, capability-negotiation trust, and delegation-chain…
Test a web app's authorization: IDOR/BOLA, missing function-level access control, privilege escalation (horizontal and vertical), and forced browsing. Use to validate OWASP A01 on an authorized target — the most prevalent web risk.
$ npx -y skills add jassics/awesome-claude-security --skill access-control-test --agent claude-codeHow it fires
How this skill gets triggered: by you, by Claude, or both.
/access-control-testContext preview
The summary Claude sees to decide when to auto-load this skill.
Test a web app's authorization: IDOR/BOLA, missing function-level access control, privilege escalation (horizontal and vertical), and forced browsing. Use to validate OWASP A01 on an authorized target — the most prevalent web risk.
name: access-control-test description: >- Test a web app's authorization: IDOR/BOLA, missing function-level access control, privilege escalation (horizontal and vertical), and forced browsing. Use to validate OWASP A01 on an authorized target — the most prevalent web risk.
Evidence on whether the app enforces authorization server-side on every object and function, tied to the authenticated user — or whether a user can reach data/actions they shouldn't.
A, User B, and an admin) with identifiable resources.
1. **IDOR / BOLA (horizontal)** — as User A, manipulate object identifiers (IDs, UUIDs, filenames, GUIDs in URLs/bodies/JWT) to access User B's resources. 2. **Function-level (vertical)** — call admin/privileged endpoints as a normal user; check hidden/undocumented functions and HTTP-method overrides. 3. **Forced browsing** — request resources/pages not linked for your role. 4. **Privilege escalation** — tamper with role/tenant claims (JWT, cookies, hidden fields, mass-assignment) to elevate. 5. **Missing ownership checks on writes** — update/delete another user's object.
1. Map roles and sensitive objects/functions. 2. Run cases as the lower-privilege identity; never use the higher account's session to "prove" access. 3. Record per case: enforced / leaked / action-performed, with request+response evidence (redact secrets). 4. Identify the gap: client-side-only checks, IDs trusted from the request, missing server-side ownership/role enforcement.
A results table: case · object/function · result · evidence · remediation (deny-by-default, server-side authz tied to identity). Confirmed issues → `security-reporting:finding` (rate high+ for cross-user data access).
Authorization must be enforced on the **server**, per object and per function, against the authenticated identity — never trust IDs, roles, or tenant hints from the client. This is the same class as API `BOLA/BFLA` — see `api-security`.
A Claude Code plugin marketplace for the full cybersecurity & GenAI-security lifecycle — from recon and threat modeling to detection engineering, GRC, and CISO-level strategy. A pentester knows which OWASP test bends a broken-access-control endpoint.
Repo: jassics/awesome-claude-security
Review agent-to-agent (A2A) / multi-agent-system trust: peer identity and authentication, message integrity, capability-negotiation trust, and delegation-chain…
Test the agent execution harness/runtime itself — LangChain/LangGraph, AutoGen, CrewAI, custom ReAct-style loops, or computer-use/browser-use agents — for…
Assess an autonomous / tool-using AI agent for security end-to-end: tool privileges, autonomy and approval boundaries, excessive agency, memory/state…
Test what an AI agent will actually do without human confirmation, including under injected-goal / prompt-injection scenarios, to validate its autonomy and…
Review the security of MCP (Model Context Protocol) servers/clients an agent uses: server trust tier, tool/resource description and result poisoning,…
Inventory the tools/functions an AI agent can call and audit their privileges, side effects, and approval requirements to find excessive-agency and…