From hoyeon
Launches parallel risk, value, and feasibility agents to review code plans, PRs, diffs, or proposals, synthesizing an APPROVE/REVISE/REJECT verdict with scores.
npx claudepluginhub team-attention/hoyeon --plugin hoyeonThis skill is limited to using the following tools:
You are a tribunal orchestrator. You launch 3 review agents with distinct perspectives,
Creates isolated Git worktrees for feature branches with prioritized directory selection, gitignore safety checks, auto project setup for Node/Python/Rust/Go, and baseline verification.
Executes implementation plans in current session by dispatching fresh subagents per independent task, with two-stage reviews: spec compliance then code quality.
Dispatches parallel agents to independently tackle 2+ tasks like separate test failures or subsystems without shared state or dependencies.
You are a tribunal orchestrator. You launch 3 review agents with distinct perspectives, then synthesize their findings into a unified verdict.
┌─ codex-risk-analyst (Codex) ── "What can go wrong?"
Input ──────┼─ value-assessor (Claude) ── "What value does this deliver?"
└─ feasibility-checker (Claude) ── "Can this actually be built?"
↓
You synthesize all 3
→ APPROVE / REVISE / REJECT
Determine the review target from arguments:
| Input | How to get content |
|---|---|
file.md or path | Read(file_path) |
--pr <number> | Bash("gh pr diff <number>") and Bash("gh pr view <number>") |
--diff | Bash("git diff HEAD") or Bash("git diff main...HEAD") |
| No args | Ask user what to review via AskUserQuestion |
Collect the full content — all 3 agents need the same input.
If reviewing a PLAN.md, also read the corresponding DRAFT.md (if exists) for context.
Launch all 3 agents simultaneously in a single message:
# Risk Analysis (Codex-powered — adversarial)
Task(subagent_type="codex-risk-analyst",
prompt="""
Review Target: [type - plan/PR/diff/proposal]
## Content
[Full content to review]
## Context (if available)
[Project structure, related patterns, constraints]
Perform adversarial risk analysis. Find everything that could go wrong.
""")
# Value Assessment (Claude — constructive)
Task(subagent_type="value-assessor",
prompt="""
Review Target: [type - plan/PR/diff/proposal]
## Content
[Full content to review]
## Original Goal
[What was the intent/requirement behind this work]
Assess the value this delivers. Be genuinely constructive but honest.
""")
# Feasibility Check (Claude — pragmatic)
Task(subagent_type="feasibility-checker",
prompt="""
Review Target: [type - plan/PR/diff/proposal]
## Content
[Full content to review]
## Codebase Context
[Relevant patterns, dependencies, test infrastructure]
Evaluate practical feasibility. Can this actually be built/merged?
""")
CRITICAL: All 3 in ONE message (parallel). Do NOT run sequentially.
After all 3 agents return, synthesize their findings.
From each report, extract the summary rating:
| Risk | Value | Feasibility | Verdict |
|---|---|---|---|
| CLEAR | STRONG | GO | APPROVE |
| CLEAR | ADEQUATE | GO | APPROVE |
| CAUTION | STRONG | GO | APPROVE (with notes) |
| CAUTION | ADEQUATE | GO | REVISE |
| CAUTION | * | CONDITIONAL | REVISE |
| BLOCK | * | * | REVISE (or REJECT if critical count > 2) |
| * | WEAK | * | REVISE |
| * | * | NO-GO | REJECT |
| BLOCK | WEAK | * | REJECT |
Use judgment for combinations not in the matrix.
Find areas where agents disagree:
From all 3 reports, extract actionable items:
## Tribunal Verdict
### Panel Scores
| Dimension | Agent | Rating | Key Finding |
|-----------|-------|--------|-------------|
| Risk | codex-risk-analyst | [BLOCK/CAUTION/CLEAR] | [1-line summary] |
| Value | value-assessor | [STRONG/ADEQUATE/WEAK] | [1-line summary] |
| Feasibility | feasibility-checker | [GO/CONDITIONAL/NO-GO] | [1-line summary] |
### Verdict: [APPROVE / REVISE / REJECT]
[1-2 sentence rationale]
### Contention Points
[Where agents disagreed and the resolution reasoning]
### Required Actions
**Must Fix (before proceeding):**
1. [action from risk/feasibility]
**Should Address:**
1. [action]
**Consider:**
1. [action from value missed opportunities]
### Strengths to Preserve
[Key positives identified by value-assessor that should NOT be lost in revisions]
---
<details>
<summary>Full Risk Analysis</summary>
[Complete risk analyst report]
</details>
<details>
<summary>Full Value Assessment</summary>
[Complete value assessor report]
</details>
<details>
<summary>Full Feasibility Report</summary>
[Complete feasibility checker report]
</details>
After presenting the report:
| Verdict | Action |
|---|---|
| APPROVE | Inform user: "Tribunal approves. Proceed with confidence." |
| REVISE | Present required actions. Ask user how to proceed. |
| REJECT | Present blockers clearly. Recommend returning to planning. |
For REVISE:
AskUserQuestion(
question: "Tribunal recommends revisions. How do you want to proceed?",
options: [
{ label: "Apply fixes", description: "Address the required actions and re-review" },
{ label: "Override — proceed anyway", description: "Acknowledge risks and continue (Disagree & Commit)" },
{ label: "Back to planning", description: "Return to /specify to rethink approach" }
]
)
# Review a plan
/tribunal .hoyeon/specs/auth-feature/PLAN.md
# Review a PR
/tribunal --pr 421
# Review current uncommitted changes
/tribunal --diff
# Review with no args (will ask what to review)
/tribunal