review-commands

name	review-commands
description	AgentSpec review commands. Use /review-commands + command. Commands: /review, /judge
license	MIT
compatibility	GitHub Copilot VS Code, GitHub Copilot cloud agent
metadata	{"version":"1.0.0","category":"commands"}

Review Commands

Invoke as:

/review-commands /review <file-or-directory>
/review-commands /judge <file> [--context "..."] [--model <openrouter-model>]
/review-commands /judge --ledger

Grounding

1. Read .github/config/grounding.md.
2. Treat this skill as the primary routing source.
3. For /review, use .github/agents/python.code-reviewer.agent.md.
4. For /judge, use .github/agents/dev.judge-agent.agent.md.
5. Load only the KB files required by the selected agent or by the reviewed domain.

/review

Use for code review, config review, migration review, and documentation review.

Execution:

1. Read the target file or directory.
2. Use the code-review stance: findings first, ordered by severity.
3. Cite concrete file paths and line numbers.
4. Include missing tests or residual risks.

/judge

Use for a second opinion on high-risk or ambiguous output.

Execution:

1. Read .github/agents/dev.judge-agent.agent.md.
2. If the request targets a file, read the file before judging.
3. If an external judge runtime is configured and OPENROUTER_API_KEY is available, run that runtime.
4. If the runtime or key is missing, stop and report the missing prerequisite instead of inventing an external verdict.
5. Write ledger entries only to .github/storage/judge-ledger.jsonl.

Comandos Disponíveis

Comando	Descrição	Arquivo	Agente
/review	Code review com findings por severidade	commands/review.md	code-reviewer
/judge	Segunda opinião via judge runtime externo	commands/judge.md	judge-agent

Constraints

• Do not use #skill: or skill: syntax.
• Do not use .claude/** paths.
• Do not initiate SDD workflow phases from this skill.