// 深度なコードレビュー // NEXT-GEN PARALLEL AGENTS

CODE REVIEW THAT DIGS DEEPER

PR Scrutiny orchestrates parallel specialist AI agents for Security, Code Quality, Blast Radius, and Intent directly inside your GitHub Pull Requests.

⚡ INSTALL ON GITHUB ◈ VIEW SOURCE
PR-SCRUTINY-ORCHESTRATOR // AGENT_DISPATCH CLICK COMMAND BELOW
TRY:
pull_request#42 > /review
------------------------------------------------------------
[SecurityAgent] Scanning diff & dependencies... WARN: Missing JWT claim validation
[QualityAgent] Analyzing complexity hotspots... PASS: Complexity within limits
[BlastRadiusAgent] Graph traversal... 3 routes & 8 controllers affected
[LLMAgent] Synthesizing summary & verification intent... DONE
------------------------------------------------------------
✔ Status: Review posted to GitHub PR #42 in 1.4s (4 parallel sub-agents).
// SPECIALIST UNITS //

PARALLEL AGENT ARCHITECTURE

🛡️

Security Agent

Detects hardcoded secrets, dangerous injection sinks, unauthenticated endpoints, and known dependency CVEs.

⚡

Quality Agent

Catches infinite loops, N+1 query patterns, cognitive complexity spikes, and missing test coverage.

🌐

Blast Radius

Builds an AST-aware import dependency graph to pinpoint exactly which downstream modules and API routes are affected.

🧠

LLM Specialist

Generates plain-English intent summaries, answers conversational diff questions, and checks architectural alignment.

// COMMAND INTERFACE //

SUMMON WITH SLASH COMMANDS

/review
Full Review Runs all 4 specialized agents in parallel and posts inline comments + summary.
/review:security
Security Scan Dedicated deep-dive on auth gaps, secret leaks, and supply-chain vulnerabilities.
/blast-radius
Impact Graph Visualizes ripple effects, affected routes, and config dependencies across the codebase.
/ask <question>
Conversational Query Ask arbitrary questions about the diff: /ask will this migration lock tables?
/summarize
Diff Summary Generates an executive explanation of code changes, reasons, and potential side-effects.
/re-review
Differential Re-Scan Re-runs analysis on updated commits to verify if previously flagged issues were resolved.
// SIGNAL HIERARCHY //

ACTIONABLE SEVERITY RATINGS

HOLD Critical blocker — critical security vulnerability or crash hazard. Blocks PR approval.
WARN High risk or subtle bug — should be carefully reviewed and addressed before merge.
SUGGEST Performance or clean code enhancement — recommended but non-blocking.
PASS Code segment verified and healthy according to specialist agents.