code-review-graph

Token-efficient code review using Tree-sitter AST graphs and MCP. Cuts AI token usage on large codebases by computing the blast radius of changes instead of reading entire codebases. Uses a SQLite graph database for structural analysis.

Quality

61%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Passed

No findings from the security scan

Fix and improve this skill with Tessl

tessl review fix ./.agents/skills/code-review-graph/SKILL.md

Quality

Content

63%Weight 40%Scale 1-5

Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.

The body is highly actionable with concrete commands and a well-checkpointed install/bootstrap flow, but it is a monolithic ~300-line document with no bundle files or external references, and it repeats when-to-use guidance across several sections.

Suggestions

Split tangential sections (Alternatives Comparison, Integration with AG Kit, Known Limitations) into separate reference files under ./references/ and link to them one level deep, so SKILL.md stays a lean overview.

De-duplicate the when-to-use criteria — keep one canonical 'When to Use vs Skip' section and remove the overlapping lists in the Bootstrap Protocol and Best Practices sections.

Replace the hedged 'illustrative' token-impact table with a single concrete example or move it to a benchmarking reference, cutting tokens spent on disclaimers.

Dimension	Reasoning	Score
Conciseness	The ~300-line body is mostly information-dense, but repeats the when-to-use criteria (frontmatter, Bootstrap Protocol, 'When to Use vs Skip', and Best Practices sections overlap), carries a hedged 'illustrative — varies by codebase' token table, and includes tangential prose (AG Kit session architecture) that could be tightened.	3 / 5
Actionability	Provides numerous concrete, copy-pasteable commands ('pipx install code-review-graph', 'code-review-graph build', 'code-review-graph rename preview --from OldClassName --to NewClassName', '/mcp' verification), with only minor gaps such as missing example outputs and full flag coverage for some subcommands.	4 / 5
Workflow Clarity	The Bootstrap Protocol and 5-step Installation give a clear sequence with explicit confirmation checkpoints ('ask the user before running code-review-graph build', 'Never install or run build without confirmation', 'Verify Integration: run /mcp and confirm'), with only minor validation gaps in the analysis workflows.	4 / 5
Progressive Disclosure	Section headers are well-organized, but at ~300 lines everything is inlined in a single file with zero external references; tangential blocks (Alternatives Comparison, Integration with AG Kit, Recommended Session Architecture) would belong in separate reference files for a skill this size.	3 / 5
	Total	14 / 20 Passed

Description

58%Weight 40%Scale 1-5

Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.

The description is specific about mechanism and distinct in niche, but relies on technical jargon for triggers and lacks an explicit 'when to use' clause in the description field itself (it is only present in when_to_use), capping completeness.

Suggestions

Add an explicit 'Use when...' clause directly in the description with concrete natural trigger phrases (e.g., 'Use when reviewing large codebases or monorepos, making multi-file changes, or trying to cut AI token costs').

Swap some jargon for natural user synonyms — add 'monorepo', 'token cost', 'refactoring preview', and 'dead code detection' so users' actual phrasings match the description.

Move one or two concrete review outcomes (risk scoring, dead-code detection) into the description so the 'what' is comprehensive rather than only mechanistic.

Dimension	Reasoning	Score
Specificity	Names several concrete actions — 'using Tree-sitter AST graphs and MCP', 'computing the blast radius of changes', 'Uses a SQLite graph database for structural analysis' — but stops short of listing the review outcomes (risk scoring, dead-code detection, refactoring preview) that appear only in the body, so it is not fully comprehensive.	4 / 5
Completeness	Has a clear 'what' (token-efficient structural code review) but no explicit 'when' trigger in the description — the 'Use when...' guidance lives only in the separate when_to_use field, so per rubric guidance completeness is capped at 3.	3 / 5
Trigger Term Quality	Contains some natural terms ('code review', 'large codebases', 'AI token usage') but mixes in technical jargon ('blast radius', 'Tree-sitter AST graphs', 'MCP', 'SQLite') and omits common user synonyms like 'monorepo', 'token cost', or 'refactoring'.	3 / 5
Distinctiveness Conflict Risk	The Tree-sitter/SQLite/blast-radius/MCP framing carves a distinct niche, but the generic 'code review' lead term carries minor overlap risk with other review-oriented skills.	4 / 5
	Total	14 / 20 Passed

Validation

87%

Warnings & errors only

Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.

Validation — 14 / 16 Passed

Validation for skill structure

Criteria	Description	Result
allowed_tools_field	'allowed-tools' contains unusual tool name(s)	Warning
frontmatter_unknown_keys	Unknown frontmatter key(s) found; consider removing or moving to metadata	Warning

	Total	14 / 16 Passed

Repository: vudovn/ag-kit
Path: .agents/skills/code-review-graph/SKILL.md
Commit: 549ba3d

Reviewed: 2 days ago

Table of Contents

Discovery Implementation Validation

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.