mcp-server-security-review
Skill0 starsChanged 13 days agoMITView on GitHub
Review Model Context Protocol (MCP) servers and client configurations — server provenance and pinning, tool description poisoning, authorization (OAuth 2.1 per the MCP spec, no token passthrough), confused-deputy risks, local server sandboxing, secret handling, and tool-call logging. Use when installing or approving an MCP server, building one, or assessing MCP use in coding assistants and agents.
1 other project uses a version of this.
Adopted by
Most recently taken up by
- howardhsieh/agent-security-skillstheir own version, first seen Oct 13, 2026
Other versions1 project uses a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 1
- howardhsieh/agent-security-skills★ 2same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository273 other files it gives its agents.
- ad-attack-path-analysisSkill
- adcs-security-reviewSkill
- ad-privileged-access-reviewSkill
- dcsync-detectionSkill
- gpo-security-reviewSkill
- kerberos-attack-detectionSkill
- agent-tool-permission-reviewSkill
- ai-incident-responseSkill
- llm-gateway-controlsSkill
- llm-security-evaluationSkill
- model-artifact-securitySkill
- prompt-injection-reviewSkill
- rag-security-reviewSkill
- api-abuse-detectionSkill
- api-authorization-testingSkill
- api-inventorySkill
- api-schema-securitySkill
- dast-baselineSkill
- sast-triageSkill
- sca-dependency-riskSkill
- secure-code-reviewSkill
- security-requirementsSkill
- threat-modelingSkill
- asset-criticality-tieringSkill
- asset-inventory-reconciliationSkill
- external-attack-surface-reviewSkill
- shadow-it-discoverySkill
- aws-credential-compromiseSkill
- aws-iam-reviewSkill
- aws-org-guardrailsSkill
- cloudtrail-investigationSkill
- guardduty-triageSkill
- s3-exposure-reviewSkill
- azure-activity-investigationSkill
- azure-rbac-reviewSkill
- conditional-access-reviewSkill
- entra-audit-investigationSkill
- entra-signin-investigationSkill
- backup-assuranceSkill
- critical-service-mappingSkill
- ransomware-recoverySkill
- recovery-exerciseSkill
- cloud-exposure-reviewSkill
- cloud-incident-responseSkill
- cloud-logging-baselineSkill
- cloud-posture-assessmentSkill
- audit-evidence-packageSkill
- continuous-control-monitoringSkill
- control-mappingSkill
- control-testing-samplingSkill
- container-incidentSkill
- dockerfile-hardeningSkill
- image-scanningSkill
- runtime-detection-falcoSkill
- db-audit-loggingSkill
- db-exfiltration-investigationSkill
- db-hardeningSkill
- db-privilege-reviewSkill
- data-classificationSkill
- data-discoverySkill
- The first 60 shown.
The file
Files
MCP Server Security Review
Threats
| Threat | Description |
| Malicious or compromised server | Local stdio servers run as your user with your files and credentials; remote servers see every tool call |
| Tool poisoning | Tool names/descriptions/outputs contain hidden instructions that steer the model (indirect prompt injection via metadata) |
| Rug pull | Server updates change tool behaviour/descriptions after approval |
| Token passthrough / confused deputy | Server forwards the client's token to downstream APIs, or uses its own broad credentials on behalf of any user |
| Excessive scopes | OAuth scopes far beyond tool needs |
| Data exfiltration | Tools that fetch URLs or send data externally |
| Session hijack | Weak session IDs on HTTP transports |
Review checklist
- Provenance: known publisher, source available, version pinned (package version/commit/container digest), dependencies scanned (
sca-dependency-risk). - Tools: read each tool's description and schema for hidden/odd instructions; minimal tool set enabled; destructive tools require client-side approval.
- Local servers: run in a sandbox/container with minimal filesystem and network access; no broad home-directory access; secrets supplied via env from a manager, not committed config.
- Remote servers: HTTPS; authorization per the MCP spec (OAuth 2.1 with PKCE, resource indicators/audience-bound tokens); server validates tokens issued for itself and does not pass client tokens through to upstream APIs; per-user authorization for upstream calls.
- Logging: tool calls and parameters logged on the client and server; alerts for unusual tools or volumes.
- Change control: re-review on version updates; monitor tool description changes (hash descriptions).
- Client config: project-level MCP configs in repositories can be injection vectors — require approval before enabling servers defined by a repo.
Output
Server | Version/digest | Transport | Tools (side effects) | Auth model | Sandbox | Findings | Decision
Related skills
agent-tool-permission-review, prompt-injection-review, sca-dependency-risk, oauth-app-review
References
- MCP specification — Security best practices: https://modelcontextprotocol.io/specification/draft/basic/security_best_practices
- MCP specification — Authorization: https://modelcontextprotocol.io/specification/draft/basic/authorization
- OWASP GenAI Security Project: https://genai.owasp.org/
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.