/synthex:review-code/synthex:review-code
BuildMulti-perspective review — Code + Security + (optional) Performance.
Invocation
Run from inside Claude Code in any project where Synthex is installed:
/synthex:review-codeSome commands accept parameters; see the source-of-truth file linked below for the full parameter list and workflow.
Lifecycle phase
Build. Parallel-executed, peer-reviewed task delivery.
Related agents
- Tech Lead · Full-stack orchestrator. Breaks tasks down, delegates to specialists, rolls up the result.
- Lead Frontend Engineer · Coordinates frontend delivery, enforces UX quality, accessibility, and design-system fidelity.
- Architect · System design, ADRs, plan-review feasibility, technology trade-offs.
- Code Reviewer · Craftsmanship review — correctness, maintainability, reuse, conventions.
- Security Reviewer · Mandatory gate. Secrets, injection, auth, supply chain, OWASP.
- Quality Engineer · Test strategy, gap analysis, behavior-driven test authoring.
- Design System · Token ownership, component governance, compliance audit.
- Technical Writer · API docs, guides, migrations, changelogs verifiable against code.
Source
The command's parameters, workflow, and sub-agent invocations are defined in markdown at bluminal/lumenai/blob/main/plugins/synthex/commands/review-code.md.
Last modified:
What it does, end-to-end
/synthex:review-code/synthex:review-code is the standalone command for reviewing a diff. It runs the same reviewer pipeline that /synthex:next-priority/synthex:next-priority uses internally during the build phase — but here it's user-facing, so you can invoke it on demand against any diff.
Each invocation:
- Resolves the diff under review. Default: the staged + unstaged changes in the current git worktree. Override via parameters to target a specific commit range, branch, or PR diff.
- Assembles the context bundle. The Context Bundle Assembler agent gathers the diff,
the surrounding code (where reviewers might need to understand callers/callees), the
project conventions (
CLAUDE.md,.eslintrc,.prettierrc), and any relevant specs undercode_review.spec_paths. - Fans out to reviewers. The mandatory pair — Code Reviewer
and Security Reviewer — runs in parallel,
along with any specialists configured under
code_review.specialistsorcode_review.reviewers. - Optionally fans out to multi-model. When the project's config opts in (per-command
multi_model: trueor via the FR-MR21 decision order based on diff risk + size), the Multi-Model Review Orchestrator also runs Claude alongside external CLI adapters (GPT, Gemini, etc.). All verdicts feed into the same consolidator. - Consolidates findings. The Findings Consolidator dedupes, preserves attribution, ranks by severity. Output is one tractable list, not three-to-five overlapping reports.
- Iterates the review loop. If findings at or above
min_severity_to_address(defaulthigh) remain, the orchestrator (Tech Lead, when invoked from /synthex:next-priority/synthex:next-priority, or you, when run standalone) addresses them and reviewers re-run. Bounded bymax_cycles(default2). - Emits a verdict. PASS / WARN / FAIL with the consolidated findings list. PASS = no findings at or above the threshold. WARN = findings exist below the threshold. FAIL = findings at or above remain after the cycle cap.
Reviewer roster
Default mandatory pair:
code_review:
reviewers:
- code-reviewer
- security-reviewerAdd specialists for performance-critical or regulated codebases:
code_review:
reviewers:
- code-reviewer
- security-reviewer
- performance-engineer
specialists:
- agent: compliance-reviewer
focus: "Regulatory compliance, audit-trail requirements"The convention: reviewers always run; specialists attach when the diff matches their
focus area (regex or path-based, defined per specialist).
What "PASS" actually means
A PASS verdict is not a guarantee the code is correct. It is a guarantee that:
- Both mandatory reviewers ran on the diff.
- All findings at or above the configured severity threshold were addressed.
- The diff matches the project conventions in
convention_sources(within the reviewer's ability to detect). - No spec violations were found in the configured
spec_paths.
It does not guarantee runtime correctness, performance, or completeness — those are
covered by tests, audits, and observation, not review. The reviewers are advisory; the
final accountability for what merges is the orchestrating agent's (and yours, for [H]
gates).
When to invoke it directly
You normally don't — /synthex:next-priority/synthex:next-priority wraps it during the build phase. Reach for /synthex:review-code/synthex:review-code standalone when:
- Running review on a non-Synthex-orchestrated change (manual edit, hotfix, contributor PR).
- Reviewing a diff that's larger than
code_review.max_diff_lines(default300) — the reviewer warns and gives lower-confidence findings, but the verdict is still useful. - Re-running review after addressing prior findings without re-engaging the full Tech Lead workflow.
Parameters
| Parameter | Description | Default | Required |
|---|---|---|---|
diff_target | What to review: a branch, commit range, or "working" for staged+unstaged | working | No |
multi_model | Force multi-model review on or off (overrides config) | Value from code_review.multi_model | No |
severity | Override min_severity_to_address for this run | Value from code_review.review_loops | No |
Common scenarios
"Diff is too large to review"
The reviewer warns when a diff exceeds max_diff_lines. You have two options:
- Split the diff into smaller PRs, then review each. This is the higher-quality path.
- Proceed with the warning and accept that findings will be less precise.
Synthex defaults to the warning rather than refusing, because real projects sometimes need to review a large change.
"All findings are below the threshold"
PASS verdict, but the consolidated findings list is non-empty. These are recorded in the
output for your review; they don't block merge but they're worth scanning. Some teams set
a quarterly cleanup cadence to address accumulated medium findings as a batch.
"Reviewers disagree on severity"
The Findings Consolidator preserves both verdicts with attribution. You decide. Disagreement is high-signal — usually it means the issue has a nuance one reviewer caught and the other didn't.