agent-tool-permission-review
Skill0 starsChanged 15 days agoMITView on GitHub
Review AI agent and copilot tool permissions for excessive agency (OWASP LLM06:2025) — tool inventory, identity used for each tool call (user-delegated vs. service), scopes, side effects, human-in-the-loop approvals, sandboxing of code execution, rate limits, and audit logging. Use when deploying an AI agent, coding assistant, or automation that lets an LLM call tools, APIs or shell commands.
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
- ai-incident-responseSkill
- llm-gateway-controlsSkill
- llm-security-evaluationSkill
- mcp-server-security-reviewSkill
- 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
We split this skill into the rules it gives an agent
1 rule. About 0 are common, 0 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈0
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 21
Minimal toolset: remove generic tools when specific ones suffice (a send_email(to=fixed-domain) beats http_request(any_url).
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Agent Tool Permission Review
Tool inventory
For every tool/function/MCP server the agent can call:
Tool | Operation(s) | Side effects (read/write/delete/send/pay/execute) | Identity used | Scope/permissions | Data reachable | Approval required? | Rate limit | Logged?
Principles
- Minimal toolset: remove generic tools when specific ones suffice (a
send_email(to=fixed-domain)beatshttp_request(any_url); aquery_orders(user)beats raw SQL). - User-delegated identity: tool calls execute with the end user's permissions (OAuth on-behalf-of, scoped tokens), not a powerful shared service account — prevents confused-deputy escalation.
- Read-only default; write/delete/send actions require explicit human approval with a faithful preview of parameters.
- Sandbox code execution: containers/microVMs without network or with egress allow-lists, no host mounts, no credentials, CPU/time limits.
- Constrain parameters server-side (validate schemas, allow-lists for recipients/paths/domains); never trust the model to self-limit.
- Budget & rate limits per session/user to bound damage and cost (LLM10 unbounded consumption).
- Audit: log each tool call (who, which tool, parameters, result status) to the SIEM; alert on anomalous sequences (bulk reads followed by external sends).
- Kill switch to disable tools/agent quickly.
Coding agents (e.g., in IDE/CLI)
Restrict shell/file permissions with allow-lists, require approval for network and destructive commands, run in containers/devcontainers for untrusted repos, don't expose production credentials in the environment, review hooks/config files (they can be injection vectors through repository content).
Test
Attempt (authorized) prompt-injection-driven misuse of each write tool (prompt-injection-review), privilege boundaries (can user A's session read user B's data via the agent?), and approval bypass (does a multi-step plan circumvent confirmation?).
Related skills
prompt-injection-review, mcp-server-security-review, trust-boundary-analysis, llm-gateway-controls
References
- OWASP LLM06:2025 Excessive Agency: https://genai.owasp.org/llmrisk/llm062025-excessive-agency/
- OWASP Agentic AI – Threats and Mitigations: https://genai.owasp.org/resource/agentic-ai-threats-and-mitigations/
- NIST AI RMF 1.0: https://www.nist.gov/itl/ai-risk-management-framework
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.

