Code Review

Metadata	Value
Category	agents
Capabilities	llm, shell
Website	https://agentos.to

Readme

Evaluate code changes against project principles and architectural direction. Reads the principles docs, active refactoring specs, and crate architecture to score diffs and catch violations at commit time.

Knowledge Base

The evaluator reads these documents before scoring:

Source	What	When to apply
core/principles.md	Engine principles — Rust is generic, templates do the work, etc.	Every Rust commit
docs/src/content/docs/principles.md	Shape principles — no counts, edges are verbs, etc.	Shape/ontology changes
docs/src/content/docs/skills.md	Skill SDK reference — how skills are built	Any skill/Python commit
core/docs/specs/refactoring/*.md	Active refactoring specs — what’s flagged for cleanup	Any commit touching flagged files
./dev.sh arch output	Live crate/module sizes, dependency chain, largest files	Structural awareness

Tools

evaluate_commit

Score a git diff against the project’s principles and refactoring specs.

Parameters:

• diff (required) — the staged git diff
• files — newline-separated list of changed file paths
• threshold — minimum passing score (default: 90)
• model — LLM model to use (default: sonnet)

Returns: { score, maxScore, pass, violations, summary }

Future Tools

• review_pr — evaluate all commits in a PR