Works with the AI Agents you already use
riskproof data leak check
Audit agent workflows and MCP tool configurations for sensitive data leaks and exposure risks.
Free
riskproof data leak check
Example session with this skill installed
Review this MCP tool definition for a Jira integration to see if it risks leaking local environment variables or customer PII during issue creation.
- Read your context and instructions
- Compiled the riskproof data leak
I have reviewed the Jira MCP tool definition and the associated issue-creation workflow. The review focused on data flow from local environment variables to the Jira create_issue endpoint.
Data Leak Review: Jira MCP Workflow
| Source | Sink | Data Class | Risk Level | Evidence |
|---|---|---|---|---|
process.env | Jira API Description | System Config | Medium | config.ts:42 passes env vars to description |
| User Input | Jira API Comments | PII | High | No redaction filter on task_details field |
| Workflow Context | Jira Labels | Internal Metadata | Low | Labels include internal project IDs |
Recommended Fixes
Implement Redaction: Add a regex-based filtering utility in jira_client.ts to intercept task_details before the API call.
Restrict Environment Access: Refactor config.ts to only export the necessary JIRA_API_KEY, removing the spread operator on process.env.
Synthetic Verification Plan
- Input: "Fix bug in system for user test@example.com using DB_PASS=secret123"
- Policy: PII and Env Vars must be redacted.
- Expectation: The Jira API payload should contain "[REDACTED]" for the email and omit the environment string.
- Method: Run the workflow in
dry-runmode and inspect the logged JSON payload.
Next steps
- Apply the redaction filter to the
jira_client.tsutility. - Audit the
config.tsfile to explicitly white-list environment variables. - Run the synthetic verification test to confirm the fix works as intended.
Connects securely to your tools. The creator never sees your data.
About this skill
Stop leaking credentials and PII through your agentic workflows. This skill provides a focused security review of your agent or MCP configurations, workflow code, and tool definitions to identify where sensitive data enters the system and where it might be exposed to unauthorized sinks.
What it does
- Map data flows by identifying sources, transitions, and external sinks within your agent's code.
- Audit tool boundaries to check if external actions or transformations lack proper data minimization or redaction.
- Identify coverage gaps where implementation details are missing or runtime evidence is unavailable.
- Generate synthetic verification plans using fake inputs to test policy expectations without risking real data.
- Propose narrow fixes for supported concerns using controls already available in your specific environment.
How it works
- Scope coverage by analyzing your selected files, tool definitions, and redacted logs to identify intended destinations and restrictions.
- Evaluate paths to separate observed events from configured behavior, checking for destination limits and approval controls.
- Draft recommendations that state the affected data class, evidence of risk, and the concrete code changes required.
- Finalize review by delivering a flow map, prioritized findings, and synthetic test cases to verify the resolution.
Frameworks & tools
Works with any agentic framework or MCP (Model Context Protocol) configuration. It inspects tool definitions, workflow logic, and configuration files without requiring external scanning services or package installations.
Why this beats prompting it yourself
Generic prompts often overlook the "degrees of separation" between an untrusted input and a sensitive tool sink. This skill follows a structured review method that forces evidence-based mapping, ensuring you don't just rely on a tool's name but verify its actual implementation and data access boundaries.
Use cases
- Reviewing a new MCP tool connection for potential PII exposure.
- Auditing a changed workflow that gains access to sensitive data sources.
- Creating a synthetic test suite to verify data redaction logic.
- Identifying missing authorization controls at external API boundaries.
Known limitations
Does not perform live network testing or execute tools. Cannot read raw credential files or private keys, requiring redacted copies for inspection. Review quality depends on provided file scope.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Security.
No live execution or network testing. Requires redacted materials for credentials. Cannot access private keys or raw customer datasets directly.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Free to use
No account required to browse
Trust & safety
Security scan
Security scan passed 7 days ago
- Free with an Agensi account
- Software
Node.jsdsh-riskproof