dr-security-review

name	dr-security-review
version	2026-03-26.1
description	Use for security-sensitive changes or reviews. Focus on injection, auth, session handling, auditability, sensitive logging, and the repo’s required adversarial test posture.

DR Security Review

Priorities

Load dr-repo-reference first for shared security defaults and component verification commands.

From AGENTS.md, treat these as first-class:

• credentials or secrets exposure
• SQL injection
• command injection
• missing auth or authz
• sensitive logging
• unsafe failover or data-loss paths

Review Checklist

When the change touches handlers, commands, DB access, sessions, tokens, or admin flows, check:

• input validation
• parameterized queries
• command construction
• auth and authorization gates
• CSRF implications
• session lifecycle
• audit logging
• secret or token leakage in logs

For non-trivial security review, record:

• Observed: exact risky patterns, boundaries, or verification output
• Inferred: security consequence supported by those facts
• Proposed: mitigation or required test
• Confidence: 0.0-1.0 with one-line reason
• Falsifier: what result would show the suspected issue is not real
• Reflection: weakest assumption and the next security check

Required Test Posture

Use adversarial tests when relevant. The repo guidance includes:

• SQL injection payload coverage
• command injection payload coverage
• NoSQL, LDAP, and XML style payload checks where applicable

Security fixes without a regression test are incomplete.

Useful Verification

Use the relevant component verification entries from dr-repo-reference.

Rules

• prefer explicit validation over assumptions
• do not weaken auth, CSRF, or audit paths for convenience
• if a change affects login, admin, registration, commands, tokens, or database queries, default to a security review

Subagent Use

For large or mixed-surface changes, use security_reviewer as a dedicated read-only pass. If the security issue is inside portal request handling, pair it with portal_explorer.

Use the model and reasoning defaults from docs/CODEX_AGENT_POLICY.md so security review remains on the strongest review path.