reviewer

name	reviewer
description	Canonical b00t reviewer capability — adversarial multi-framework code/datum review. Shared across all harnesses (Claude Code, opencode, Hermes, b00t-cli). Loaded via: b00t learn reviewer, or by harness-specific role supplements.
version	1.0.0
allowed-tools	All
capability_type	skill
harness_targets	["claude-code","opencode","hermes","b00t-cli"]

What This Skill Does

Canonical reviewer capability for the b00t hive. Applies MECE structural decomposition, TRIZ contradiction analysis, and Eureka insight search to code, datums, and system configs. Produces machine-parseable verdicts (APPROVE/REQUEST_CHANGES) suitable for pre-commit gates, PR review, and governance enforcement.

Canonical Across Harnesses

Harness	How It Loads	File
b00t-cli	b00t learn reviewer	This SKILL.md
Claude Code	b00t whoami --role=reviewer	AGENTS/--role=reviewer.md (thin wrapper)
opencode	Native skill load	This SKILL.md
Hermes	Skill bundle	references/ this directory

Review Protocol

1. system-normal() — precriteria gate (git stash, conflicts, branch, submodules)
2. MECE — structural decomposition, boundary issues
3. TRIZ — contradiction analysis, unnecessary trade-offs
4. Eureka — pattern detection, simplification opportunities
5. Synthesize — cross-framework triangulation
6. Verdict — APPROVE or REQUEST_CHANGES

Output Contract (STRICT — machine parsed)

VERDICT: APPROVE

or

VERDICT: REQUEST_CHANGES

Preceded by brief justification (1-3 lines). No markdown after VERDICT.

RHAI Scripts

Script	Purpose
system-normal.rhai	Precriteria gate — git stash, conflicts, branch, submodules
autoexec.rhai	Autonomous review loop — poll queue → dispatch → post → repeat

Dependencies

• _b00t_/reviewer.role.toml — role datum with capabilities, sub-agents, compliance rules
• _b00t_/multi-framework-review.datum.toml — 7-phase state machine
• _b00t_/datum-review.datum.toml — datum-specific review skill
• _b00t_/pr-risk-scorer.datum.toml — PR risk tier routing
• b00t-cli installed (for b00t script run, b00t task list, b00t agent delegate)
• RHAI runtime (via b00t script eval)

Compliance Rules

• Always run system-normal() before review operations
• Guard violations are ALWAYS REQUEST_CHANGES
• Scope drift is WARNING, not FAIL
• Never output markdown after VERDICT line
• Report phygital-twin status per session

Version

1.0.0 — Initial canonical reviewer skill. Replaces scattered role supplements with single source of truth. All harnesses reference this canonical capability.