Run a full review¶

Paths in this document are relative to the Zolletta-MetaSkill project root.

Run a comprehensive code review that combines all available review skills in parallel. The review subcommand is the orchestrator — it reads the project language from settings.json, runs general and language-specific skills in parallel batches, and aggregates the results into a single summary report.

Prerequisites¶

A project set up with /zolletta-metaskill setup (.zolletta-metaskill/settings.json must exist)

Steps¶

Step 1 — Invoke the review¶

/zolletta-metaskill review

The orchestrator reads language from settings.json and determines which skills to run.

Step 2 — Parallel skill execution¶

The orchestrator runs general skills (patterns, documentor, external-review) always, and language-specific skills when the language matches. See subcommands.md for the full skill table.

Each skill runs as a subagent and writes its findings to a markdown report in the timestamped report folder.

Step 3 — Aggregated report¶

The orchestrator collects all sub-skill reports and creates:

SUMMARY.md — overall grade, grades by area, top strengths and weaknesses, priority items
TODO.md — aggregated action items from all skills, sorted by priority

The reports are saved to .zolletta-metaskill/reports/<timestamp>/.

Review mode¶

Follows review mode — read-only, two-bucket classification, no fixes applied.

Configuration¶

Setting	Location	Default	Description
`language`	`.zolletta-metaskill/settings.json`	(detected)	Determines which language-specific skills run
`reports_dir`	`.zolletta-metaskill/settings.json`	`.zolletta-metaskill/reports`	Where report folders are created
`external_review_model`	`.zolletta-metaskill/settings.json`	`swe`	Model for the external review sub-skill