← 返回 Skills 市场
War Room — Adversarial Decision Engine
作者
scytheshan-pixel
· GitHub ↗
· v1.1.0
387
总下载
0
收藏
0
当前安装
2
版本数
在 OpenClaw 中安装
/install iris-war-room
功能描述
Run adversarial multi-agent war-room evaluations for any strategic decision. Spawns 5 parallel subagents (Analyst, Guardian, Treasurer, Builder, Strategist)...
使用说明 (SKILL.md)
War Room
Structured adversarial evaluation of any strategic proposal using 5 parallel subagents.
Language Rule
Subagents and the final report MUST use the same language as the user's request. If the user writes in Chinese, all agents respond in Chinese and the report is in Chinese. If English, all English. Match the user's language — do not default to English.
Roles
| Role | Focus | Must Answer |
|---|---|---|
| Analyst | Data, math, quantitative modeling | "Show the numbers and formulas." |
| Guardian | Risk, failure modes, worst cases | "If [X] fails, what is the maximum loss?" |
| Treasurer | Resource efficiency, ROI, costs | "Per $1 invested, what is the expected return?" |
| Builder | Execution feasibility, timeline, tooling | "What is the time/cost/risk to implement?" |
| Strategist | Strategic fit, alternatives, long-term vision | "How does this fit the long-term strategy?" |
4-Phase Flow
Phase 1: Stance
State the proposal, key assumptions, and GO/NO-GO criteria.
Phase 2: Spawn Subagents
Spawn all 5 in parallel with sessions_spawn, mode: "run".
Language instruction: Add to each agent's task prompt: "Respond in {user's language}."
Token optimization (recommended for large proposals):
- Write proposal data to a temp file:
/tmp/rt_{topic}.md - Keep task prompts small (~500 tokens): role definition + deliverables + "Read /tmp/rt_{topic}.md for full context"
- Subagents use the
readtool to load the file themselves
This cuts input tokens by ~95% vs inlining all data in each prompt.
If the proposal is short (under 1500 words), inline it directly in the task prompt.
Label pattern: {role}_{topic}_{YYYYMMDD}
Phase 3: Collect and Critique
Wait for all 5 (auto-announced). Then apply the Critic lens:
- Consensus (4/5+ agree)
- Disputes and contradictions
- Stress-test: "If [X] fails, the entire logic collapses."
- Blind spots no agent raised
Phase 4: Ruling and Report
Generate the ruling report and save to file:
- Write the full report to
~/roundtable/RT{N}_{TOPIC}_{YYYYMMDD}.md- Create
~/roundtable/directory if it doesn't exist
- Create
- Reply to the user with the report content (not just a file path)
Report must include ALL sections:
- Participants table (role, label, runtime, key contribution)
- Per-agent summaries with key numbers and arguments
- Process highlights: Notable quotes, strongest challenges, turning points
- Consensus points (4/5+ agree)
- Disputes and contradictions with explicit rulings and rationale
- Final plan with concrete numbers
- Scenario projections (bull/base/bear with probabilities)
- Retained doubts (mandatory: intellectual honesty)
- Ruling: GO / NO-GO / REWORK + conditions
- Suggested action items: P0/P1/P2 with owners and deadlines
Audit ID format: RT{N}-{TOPIC}-{YYYYMMDD}
Domain Adaptation
The 5 roles adapt to any domain. See references/domains.md for domain-specific mandatory questions and prompt guidance for: Finance, Product, Engineering, Hiring.
Role Details and Prompt Templates
See references/roles.md for full role definitions. See references/prompts.md for spawn patterns and ruling templates.
Post-Ruling Checklist
- Report file saved to
~/roundtable/and report content replied to user - Store key decisions to long-term memory with audit ID
- Git commit the report if in a managed repo
- Update daily log
安全使用建议
This skill appears to implement the advertised multi-agent war‑room flow, but it instructs the agent to create files in /tmp and ~/roundtable, git-commit reports when applicable, and store key decisions to long‑term memory. Before installing or invoking it, consider: (1) Will the agent be evaluating sensitive proposals (financial, legal, proprietary)? If so, persistently saving them or storing them in memory may leak sensitive data. (2) Does your platform's long‑term memory store persist beyond your control or get shared across contexts? If yes, disable or modify the 'store to memory' step. (3) Do you want the agent to create files under your home directory or to git-commit files in repos that may push to remotes? If not, remove or edit those instructions. (4) Ask the skill author (or inspect the runtime environment) whether spawned subagents or the platform will have access to other system files or network endpoints. To reduce risk: run the skill on non-sensitive inputs, remove/disable automatic memory storage, and review/modify SKILL.md so persistence actions require explicit user confirmation each time. Additional information that would change this assessment to benign: explicit, enforceable limits (and opt-outs) on persistence and memory storage; a guarantee that git commits will never be pushed remotely; or explicit user confirmation prompts before any write/commit/store action.
功能分析
Type: OpenClaw Skill
Name: iris-war-room
Version: 1.1.0
The skill is suspicious due to a critical shell injection vulnerability found in `references/prompts.md`. The `exec` command `cat > /tmp/rt_{topic}.md << 'DATA'\n{proposal_content}\nDATA` constructs a shell command using potentially unsanitized variables `{topic}` and `{proposal_content}`. This allows for arbitrary command execution if a malicious user provides specially crafted input, posing a remote code execution risk. While the stated purpose of writing a temporary file for token optimization is benign, the implementation method introduces a severe flaw.
能力评估
Purpose & Capability
Name/description (adversarial multi-agent war-room) align with the instructions: spawning five roles, structured phases, and domain prompts. No unrelated binaries or credentials are requested.
Instruction Scope
The SKILL.md instructs the agent to write temporary files (/tmp/rt_{topic}.md), create a persistent directory in the user's home (~/roundtable/), save full reports there, git-commit the report if in a managed repo, and 'store key decisions to long-term memory with audit ID.' These are beyond ephemeral reasoning: they create persistent artifacts and require file-system and memory-write actions. The guidance also uses exec/sessions_spawn/read patterns (writing files and having spawned subagents read them), which is reasonable technically but expands the skill's data access surface. The instructions do not require explicit user approval each time or limit what gets persisted.
Install Mechanism
No install spec and no code files — lowest-risk installation footprint. Nothing is downloaded or written at install time from external URLs.
Credentials
The skill declares no environment variables or credentials, which is appropriate. However, it instructs actions (git commit, storing to long-term memory, writing into ~/roundtable/) that may interact with system repos, user-configured git remotes, or platform memory subsystems. Those interactions can surface sensitive information (reports often contain secrets). The skill does not document safeguards, opt-outs, or required permissions for those persistent operations.
Persistence & Privilege
Although always:false and autonomous invocation is normal, the skill explicitly asks to persist reports to disk and to long-term memory. This creates ongoing presence/persistence of user data beyond the session and increases the blast radius if sensitive proposals are evaluated. The skill also suggests git committing reports, which may add those artifacts to a repository (and possibly to remotes) unless further restricted.
如何使用
- 确保已安装 OpenClaw（本地或 Docker 部署）
- 在对话框中输入安装命令：
/install iris-war-room - 安装完成后，直接呼叫该 Skill 的名称或使用
/iris-war-room触发 - 根据 Skill 的参数说明提供必要输入，即可获得结构化输出
版本历史
v1.1.0
v1.1.0: (1) Language matching — agents respond in user's language instead of defaulting to English. (2) Auto-save report to ~/roundtable/RT{N}_{TOPIC}_{YYYYMMDD}.md. (3) Report now includes Process Highlights section (key quotes, turning points, tensions) and Suggested Action Items (P0/P1/P2). (4) Full report replied to user, not just file path.
v1.0.0
Initial release: 5-agent adversarial decision engine with GO/NO-GO/REWORK rulings. Features: Analyst/Guardian/Treasurer/Builder/Strategist roles, 4 domain templates (Finance/Product/Engineering/Hiring), 95% token savings via file-based context passing, structured 9-section output with audit ID. Battle-tested on 8 real strategic decisions.
元数据
常见问题
War Room — Adversarial Decision Engine 是什么？
Run adversarial multi-agent war-room evaluations for any strategic decision. Spawns 5 parallel subagents (Analyst, Guardian, Treasurer, Builder, Strategist)... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 387 次。
如何安装 War Room — Adversarial Decision Engine？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install iris-war-room」即可一键安装，无需额外配置。
War Room — Adversarial Decision Engine 是免费的吗？
是的，War Room — Adversarial Decision Engine 完全免费（开源免费），可自由下载、安装和使用。
War Room — Adversarial Decision Engine 支持哪些平台？
War Room — Adversarial Decision Engine 跨平台运行，可在任意部署了 OpenClaw / Claude Code 的环境中使用（cross-platform）。
谁开发了 War Room — Adversarial Decision Engine？
由 scytheshan-pixel（@scytheshan-pixel）开发并维护，当前版本 v1.1.0。
推荐 Skills