sdd-review

name	sdd-review
namespace	true
keyword-tags	["injection","malicious-code","source-verification","citation","rtc","return-to-context","supply-chain","secrets","security-review"]
description	Phase 5 (Review) namespace meta-skill — injection, malicious-code, citations, RTC.

sdd-review (namespace meta-skill)

Purpose: lightweight router for Phase 5 review work.

When to Use

• Reviewing a feature before Gate 3 or Gate 4.
• The user mentions security, supply chain, citations, or reasoning quality.
• The diff includes external dependencies, copy-pasted snippets, or unverified claims.

Routed Sub-Skills

Trigger keywords	Sub-skill	Purpose
injection, prompt injection, untrusted input	(uses injection-scan instruction)	Injection scanning is in injection-scan.instructions.md
malicious, obfuscated, eval, dynamic load	malicious-code-detection	Detect malicious code patterns
supply-chain, dependency, lockfile, transitive	supply-chain-risk	Supply-chain risk assessment
secret, credential, api key, token	secrets-scan	Scan for secret leakage
citation, source, claim, external knowledge	source-citation-check / source-verification	Verify cited sources
rtc, return-to-context, reasoning	(uses rtc-reasoning instruction)	RTC reasoning is in rtc-reasoning.instructions.md
stub, placeholder, TODO, not implemented	stub-scan	Detect stub/placeholder code left in the diff (Wave 24)
red-team review, adversarial review	red-team-spec	Adversarial review

Invocation Guidance

1. Run secrets-scan and injection-scan on every diff — non-negotiable.
2. Run supply-chain-risk whenever dependencies change.
3. Run source-citation-check when external claims appear in the diff or spec.

Boundary

• Never approve a Gate without running the mandatory scans.
• Never silently downgrade a CRITICAL finding — escalate per escalation-protocol.