Skip to content
Security
Command

/deviation-request

Draft deviation requests, risk acceptances, or false positive justifications

BOOST
From plugin
claude-grc-plugin
18424 skills1 agent24 commands
Install
> /plugin marketplace add mlunato47/claude-grc-plugin
> /plugin install grc@mlunato47

How it fires

How this command gets triggered: by you, by Claude, or both.

  • Fires itselfClaude auto-loads it when your prompt matches the work.
  • You can call itInvoke it directly when you want it.
  • Slash command/deviation-request

Context preview

What this command does when you run it.

Draft deviation requests, risk acceptances, or false positive justifications

Command definition

deviation-request.md
description: "Draft deviation requests, risk acceptances, or false positive justifications"

/grc:deviation-request

Draft deviation requests, risk acceptances, or false positive justifications.

Usage

/grc:deviation-request [framework] [control-id] [reason-type]

Arguments

  • **framework**: The compliance framework
  • **control-id**: The specific control ID requiring deviation
  • **reason-type**: The type of deviation. Accepts:
  • `operational` — Operational requirement prevents full implementation
  • `false-positive` — Scanner finding is a false positive
  • `risk-adjustment` — Risk is mitigated by compensating controls
  • `vendor-dependency` — Remediation depends on vendor action
  • `technology-constraint` — Technical limitation prevents implementation
  • `temporary` — Temporary deviation with planned remediation

Examples

/grc:deviation-request fedramp si-2 vendor-dependency
/grc:deviation-request fedramp ac-7 operational
/grc:deviation-request pci 6.3.3 false-positive
/grc:deviation-request nist sc-28 technology-constraint
/grc:deviation-request fedramp ra-5 temporary

Behavior

When invoked:

1. **Identify the framework, control, and reason type** from arguments.

2. **Read the relevant framework reference** and **POA&M management reference** (`conmon/poam-management.md`).

3. **Draft a deviation request** that includes all required elements:

a. **Finding identification**:

  • POA&M ID / Weakness ID
  • Control ID and title
  • Finding source (scan tool, assessment, etc.)
  • Original severity and risk rating

b. **Deviation justification** (varies by reason type):

  • **Operational**: Why the control cannot be fully implemented, what business process depends on the exception, impact of full implementation
  • **False positive**: Technical evidence that the finding is not a true vulnerability, scanner configuration details, manual verification results
  • **Risk adjustment**: Compensating controls in place, residual risk calculation, why compensating controls adequately mitigate the risk
  • **Vendor dependency**: Vendor ticket/case number, vendor acknowledgment, expected vendor timeline, interim mitigations
  • **Technology constraint**: Technical limitation details, alternatives evaluated, compensating controls
  • **Temporary**: Planned remediation date, milestones, interim mitigations

c. **Risk assessment**:

  • Adjusted risk rating with justification
  • Impact if deviation is not approved
  • Compensating controls (if any)
  • Residual risk statement

d. **Approval chain**:

  • ISSO recommendation
  • ISSM review
  • AO approval (for FedRAMP)
  • Expiration date for deviation

4. **If no arguments provided**, ask for framework, control ID, and reason type.

Output Format

## Deviation Request

### Finding Information
- **POA&M ID**: [To be assigned]
- **Control**: [Framework] [Control-ID] — [Title]
- **Source**: [Scan/Assessment/Incident]
- **Original Severity**: [Critical/High/Moderate/Low]
- **Detection Date**: [Date]

### Deviation Type
[Operational Requirement / False Positive / Risk Adjustment / Vendor Dependency / Technology Constraint / Temporary]

### Justification
[Detailed justification based on deviation type]

### Compensating Controls
| Control | Description | Risk Reduction |
|---------|-------------|---------------|

### Risk Assessment
- **Original Risk**: [Rating]
- **Adjusted Risk**: [Rating]
- **Residual Risk**: [Statement]
- **Risk Acceptance Rationale**: [Why this residual risk is acceptable]

### Remediation Plan (if temporary/vendor)
| Milestone | Target Date | Owner |
|-----------|-------------|-------|

### Approval
- **Requested by**: [Name, Role]
- **ISSO Recommendation**: [Approve/Deny]
- **ISSM Review**: [Approve/Deny]
- **AO Decision**: [Approve/Deny]
- **Expiration Date**: [Date — typically 1 year max]

### Supporting Evidence
- [Attachment 1: Description]
- [Attachment 2: Description]

Notes

  • FedRAMP deviation requests have specific submission requirements through FedRAMP (legacy Rev5 process). Under CR26 VDR/VER there is no deviation-request path: items not remediated within 192 days become Accepted Vulnerabilities with written justification, listed in each quarterly Ongoing Certification Report.
  • False positive deviation requests require technical evidence (not just an assertion).
  • Operational requirement deviations should be rare and well-justified — auditors scrutinize these heavily.
  • All deviations should have an expiration date and be reviewed at least annually.
  • Compensating controls must demonstrably reduce the risk to an acceptable level.
Read more
Ships withclaude-grc-plugin

A plugin that turns your AI coding agent into a senior GRC (Governance, Risk, and Compliance) analyst. 76+ reference files covering 18 frameworks, 24 slash commands, and deep domain knowledge for federal and commercial compliance work.

Get the whole plugin

Other commands on claude-grc-plugin.