From rad-code-review
Review my code, code review, is this ready to ship, check for bugs, security audit, review this PR, pre-merge check, is this safe to deploy, check code quality. Blame-aware diff scoping, 3-role adversarial review, AI slop detection (14 patterns including a mechanical hallucinated-imports validator), framework IDOR (6 frameworks), WCAG 2.2, performance heuristics, severity-ranked CR-NNN findings, fingerprint-based history comparison, optional fix application. Parallel tool calls, JSON-first subagent output, compaction-safe checkpointing, non-interactive mode for agents/CI, cross-model adversarial pass via --adversarial-model.
How this skill is triggered — by the user, by Claude, or both
Slash command
/rad-code-review:rad-code-review [repo|diff|commit] [--since commit] [--strictness mvp|production|public] [--model opus|sonnet|haiku] [--adversarial-model name] [--non-interactive] [--resume RUN-ID] [--fix blockers|critical-major|IDs][repo|diff|commit] [--since commit] [--strictness mvp|production|public] [--model opus|sonnet|haiku] [--adversarial-model name] [--non-interactive] [--resume RUN-ID] [--fix blockers|critical-major|IDs]This skill is limited to the following tools:
The summary Claude sees in its skill listing — used to decide when to auto-load this skill
**Cross-model note.** Defaults to **Opus** for the primary review (best reasoning for the adversarial protocol + severity calibration). **Sonnet** is a first-class fallback — set `--model sonnet` for cost-sensitive PR scans. **Haiku** only for narrow blame-aware diffs with `--local-only`. The adversarial pass is self-adversarial (same model) by default; pass `--adversarial-model <name>` for a c...
LICENSEREADME.mdROADMAP.mdproject-types/api.mdproject-types/chrome-extension.mdproject-types/cli-tool.mdproject-types/electron-app.mdproject-types/library.mdproject-types/mobile-app.mdproject-types/saas.mdproject-types/web-app.mdreferences/adversarial-protocol.mdreferences/ai-slop-patterns.mdreferences/performance-heuristics.mdreferences/release-readiness.mdreferences/security-checklist.mdreferences/severity-model.mdreferences/subagent-prompts/adversarial-review.mdreferences/subagent-prompts/primary-review.mdreferences/subagent-prompts/self-adversarial-review.mdCross-model note. Defaults to Opus for the primary review (best reasoning for the adversarial protocol + severity calibration). Sonnet is a first-class fallback — set --model sonnet for cost-sensitive PR scans. Haiku only for narrow blame-aware diffs with --local-only. The adversarial pass is self-adversarial (same model) by default; pass --adversarial-model <name> for a cross-model challenge pass.
Naming. Finding IDs are CR-NNN (v5.0 — short to type and reference; previously RADCR-NNN). The config file (.radcrconfig.yml) and history/state directories (.radcr/history/, .radcr/state/) keep the longer radcr prefix — renaming those would break existing per-repo state for zero readability gain: you type finding IDs, not paths. Users with a .ucrconfig.yml from the oldest versions should rename it to .radcrconfig.yml and .ucr/ to .radcr/. See README.md for details.
v5.0 differentiators (new):
CR-NNN finding IDs — short to type and reference (previously RADCR-NNN)--engine claude|codex|both removed (Codex execution was never implemented); --adversarial-model provides the real cross-model pass.radcr/history/ only — no more loose root-level report files (root copy on explicit request)v3.0 differentiators (retained):
--model override--resume — compaction-safe state writes after Steps 5, 7, 9--non-interactive — agent/CI callers skip the findings menu and get structured returnreferences/subagent-prompts/.radcrconfig.yml accepted-risk expiry enforcement — stale entries re-evaluated, not silently suppressedv2.x differentiators (retained):
diff and commit scopes only flag issues on changed lines by default--since <commit> reviews all changes since a specific commitOrchestrator role: Parse arguments, compute diff scope, gather user choices, detect project context, spawn review subagents with annotated diff context, handle checkpoints and adversarial passes, offer fixes, assemble final report.
Three report roles:
Review dimensions: Functional correctness, security, AI slop detection, architecture, tests, performance, UI/UX, accessibility, release readiness, documentation, dependencies, privacy/secrets handling.
<execution_context> Load these files NOW before proceeding:
Scope options: repo | diff | commit | tree
repo — review all files in the repository (full scan, no blame filtering)diff — review staged + unstaged changes only (blame-aware by default)commit — review files changed in HEAD commit only (blame-aware by default)tree — review uncommitted working tree changes only (full scan of changed files)Incremental review: --since
Scan mode:
--full-scan — override blame-aware default, flag all issues regardless of authorshipStrictness: mvp | production | public (default: production)
mvp — focus on functional correctness, critical security, and stated goalsproduction — full review across all dimensionspublic — production + open-source readiness, public scrutiny resilience, trust signalsAdversarial pass: self-adversarial by default (same model challenges its own
findings); --adversarial-model <name> switches to a cross-model pass (a different
model family does the challenge). v5.0 removed the old --engine claude|codex|both
flag — it implied Codex execution that was never implemented; if a caller passes
--engine, say so and map both → cross-model adversarial on Opus.
Connectivity: --local-only (default: internet-enabled) Fix mode: --fix blockers | --fix critical-major | --fix id1,id2,...
Model selection (v3.0):
--model opus (default) — Opus primary review--model sonnet — Sonnet for cost-sensitive reviews--model haiku — Haiku only for narrow blame-aware + --local-only scopes--adversarial-model <name> — override adversarial-pass model separatelyNon-interactive mode (v3.0):
--non-interactive — skip the findings menu, return findings + verdict + report path. Used by the code-reviewer agent, /loop sessions, and CI.Resume (v3.0):
--resume <run-id> — rehydrate mid-review state from .radcr/state/<run-id>.json after compaction or interruption. Run IDs are logged at the start of each run.Project config: .radcrconfig.yml (if present in repo root) History: .radcr/history/{YYYY-MM-DD}-{HHmmss}-{scope}-{strictness}.md (previous review reports) State: .radcr/state/{run-id}.json (checkpoints for --resume)
Execute the orchestrate-review workflow from ${CLAUDE_SKILL_DIR}/workflows/orchestrate-review.md end-to-end.Preserve all workflow gates, user checkpoints, and subagent boundaries.
<critical_rules>
<success_criteria>
Provides CDSS development patterns for drug interaction checking, dose validation, clinical scoring (NEWS2, qSOFA), and alert classification integrated into EMR workflows.
npx claudepluginhub radorigin-llc/rad-claude-skills --plugin rad-code-review