security-review

Security Review

Conduct a focused security review using the repository threat model plus direct code inspection.

Prerequisites

• A git repository to review
• A threat model at .agent-security/threat-model.md

If the threat model is missing, generate it first with the threat-model skill.

Workflow

1. Confirm Review Scope

Determine whether this is:

• A PR review
• A staged-change review
• A branch comparison
• A full repository audit

2. Load the Threat Model

Read:

• .agent-security/threat-model.md
• .agent-security/security-config.json if present

3. Inspect the Relevant Code

Review changed files and related code paths for STRIDE issues:

• Spoofing
• Tampering
• Repudiation
• Information disclosure
• Denial of service
• Elevation of privilege

4. Validate Findings

For each finding, assess:

• Reachability
• Attacker control
• Existing mitigations
• Real exploitability

5. Produce Structured Output

Write validated results to:

• .agent-security/validated-findings.json

Then summarize:

• Confirmed findings
• Rejected findings
• Severity distribution
• Concrete remediation priorities

Gotchas

• Do not report theoretical issues without a reachability check.
• Threat-model alignment matters: review findings in system context, not in isolation.
• If a suspicious pattern is framework-mitigated, say so explicitly.