review

Engineering Community

Skill notes

Quick Start

Natural language works best. All of these trigger the skill: What You Say · What Happens /review · Review uncommitted changes (standard profile) /review security focus · Security-focused deep review /review ux focus · UX and accessibility review

Prerequisites

At least one external model CLI is recommended for multi-model consensus. Without external CLIs, Claude performs a solo review with reduced confidence. Works without external CLIs (solo mode with reduced confidence), but multi-model consensus requires at least one. See config.md for adding models and full profile definitions.

Review Profiles

Profile · Domains Reviewed · Min Models · Passes · Best For quick · Bugs & Logic, Maintainability · 1 + Claude · 1 · Small changes, typo fixes standard · All 8 code domains · 2 + Claude · 1-2 · Default for most work deep · All 8 code domains · All available · 2-3 · Critical features, releases security · Security, Robustness, Bugs & Logic · All available · 2 · Auth, APIs, data handling

Phase 1: Scope & Context

1a. Determine what to review: If the user specified files, a PR number, or a document path, use that instead. 1b. Build the reviewable file list: Include: source code (.ts, .tsx, .js, .jsx, .py, .go, .rs, .java, .rb, .sql, .c, .cpp, .h, .cs, .php, .swift, .kt, .vue, .svelte)

Phase 2: Parallel External Reviews

Run each external model review simultaneously using run_in_background: true on the Bash tool. For each model, construct a review prompt from references/prompts.md containing: The diff or file contents (inline — external models may lack workspace access) The active domain criteria from references/domains.md The structured output format specification

Fluxly profile Runtime and permissions are inferred by Fluxly static checks, not author declarations or runtime test results.
Domain
Engineering
Author-declared agents
No explicit declaration found; this is not inferred or tested compatibility
Static check
88 / 100 · heuristic scan, not runtime safety proof
Runtime evidence
Not run; no isolated runtime test evidence is currently recorded
Author / version / license
@tomevault-io · no license declared
Token usage
Moderate
Setup complexity
Guided setup
External API key
Not required
Operating systems
Windows
Runtime requirements
No explicit requirements detected
Detected file/system behavior
  • Read-only
  • Write / modify
  • Shell exec
Network behavior
Local-only
Install commands
None (reference only)

Heads up: Tool permissions are not declared; actual access depends on the runtime.

01
Phase 1: Scope & Context

1a. Determine what to review: If the user specified files, a PR number, or a document path, use that instead. 1b. Build the reviewable file list: Include: source code (`.ts`, `.tsx`, `.js`, `.jsx`, `.py`, `.go`, `.rs`, `.java`, `.rb`, `.sql`, `.c`, `.cpp`, `.h`, `.cs`, `.php`, `.swift`, `.kt`, `.vue`, `.svelte`)

02
Phase 2: Parallel External Reviews

Run each external model review simultaneously using `run_in_background: true` on the Bash tool. For each model, construct a review prompt from `references/prompts.md` containing: The diff or file contents (inline — external models may lack workspace access) The active domain criteria from `references/domains.md` The structured output format specification

03
Phase 3: Claude Self-Review

While external models run, Claude performs its own independent review: Read every changed file completely (not just the diff — full file context matters) Review against all active domains using criteria from `references/domains.md` Leverage advantages external models lack: Full codebase context (imports, callers, related files)

04
Phase 4: Consensus Synthesis

After all reviews complete, synthesize using the algorithm from `references/synthesis.md`: 4a. Parse and normalize findings: Extract findings from each model's output Normalize to common format: `{severity, domain, file, line, title, description, fix}` 4b. Cross-reference findings:

05
Phase 5: Generate Consensus Report

Output a structured report (template in `references/synthesis.md`):

06
Phase 6: Fix & Re-review Loop

If verdict is not APPROVED: Present findings to user with prioritized fix recommendations After user approves, implement fixes starting with Critical findings Re-run external reviews on changed files only (incremental) Synthesize new findings, compare with previous iteration

07
Phase 7: Final Certification

When all models approve:

Discussion

Powered by GitHub Discussions. Sign in with GitHub to comment, react, or subscribe.