CodeRabbit Security Controls

Overview

Use documented CodeRabbit Security and tools while retaining independent scanners and human review. AI findings cannot prove absence of vulnerabilities.

Prerequisites

Identify the CodeRabbit organization, Git provider, repository, plan, and accountable owner.
Read references/official-docs.md and re-check any time-sensitive contract before execution.
Use synthetic or read-only evidence until the approval boundary is satisfied.
Preserve the repository's independent CI, security, and human-review requirements.

Current Contract

CodeRabbit documents repo security scanning and security Change Stack views.
The tool catalog includes static, secret, dependency, and infrastructure analyzers.
CodeRabbit can ingest GitHub Checks.
SkillSpector scans agent skills and MCP configuration.

Authentication

Treat Git-provider sessions, CodeRabbit web sessions, CLI credentials, and CodeRabbit API keys as separate credentials. Use only an already-approved session or secret-manager reference, never print a secret, and do not place credentials in .coderabbit.yaml, source files, logs, or deliverables.

Instructions

Classify threat model, sensitive paths, languages, builds, checks, and requirements.
Map risks to supported tools, guidance, external gates, or manual controls.
Configure supported layers and retain provider security checks.
Test safe seeded fixtures, record gaps, and define escalation.

Tool Discipline

Use Glob to locate candidate configuration and evidence files without widening scope.
Use Grep to find relevant fields, commands, identifiers, and stale claims.
Use Read to inspect the smallest required files and authoritative evidence.
Use Write only for a new approved local draft or evidence artifact.
Use Edit only for a bounded approved change whose rollback is known.
Do not use these file tools as a substitute for authenticated CodeRabbit or provider operations.

Approval Boundaries

Require security approval before disabling tools, excluding sensitive paths, or accepting high-risk findings. Keep analysis and drafts local until approval is explicit, and record who approved the action and its scope.

Output

A threat-control matrix, patch, independent-gate inventory, fixture evidence, gaps, and escalation. Include source dates, unknowns, and the exact boundary between observed fact and recommendation.

Error Handling

Condition	Response
Current contract is unclear or docs disagree	Stop mutation, cite both sources, and request owner resolution.
Required access or approval is missing	Produce a draft and evidence plan only.
Validation or pilot behavior differs from expectation	Restore the prior state and retain the failed evidence.
Output contains secrets or private code	Stop, quarantine the artifact, redact it, and notify the data owner.

Examples

Example 1

Enable secret and IaC tools while keeping provider scanning.

Example 2

Test SkillSpector with a synthetic unsafe MCP fixture.

Validation

Confirm every claim against the dated sources in references/official-docs.md.
Verify the requested scope, owner, approval, happy path, failure path, and rollback.
Re-read the effective configuration or provider state after any approved change.
Report unsupported fields, undocumented endpoints, and unverified assumptions as failures.

Resources

Official documentation and contract notes
Re-check the dated contract before any live operation.
Treat unresolved or changed vendor behavior as a stop condition.

coderabbit-security-basics