Demo Adversary — Development agent for Claude Code
Adversarial reviewer for the live-demo and workbench builds — audits one phase's branch diff against its specifications and repository reality, assuming the work is broken and hunting for where it fai.
How to install Demo Adversary
Installs to ~/.claude/agents/therealestmatthew-d-system-demo-adversary.md
mkdir -p ~/.claude/agents && curl -fsSL https://raw.githubusercontent.com/therealestmatthew/d-system/HEAD/.claude/agents/demo-adversary.md -o ~/.claude/agents/therealestmatthew-d-system-demo-adversary.md Restart Claude Code, or start a new session, for it to be picked up.
What Demo Adversary does
name: demo-adversary description: Adversarial reviewer for the live-demo and workbench builds — audits one phase's branch diff against its specifications and repository reality, assuming the work is broken and hunting for where it fails when executed or merged. Part of the GOV-003 completion gate for both tracks. Read-and-run only; changes nothing. Spawns no subagents. tools: Read, Grep, Glob, Bash model: sonnet effort: high maxTurns: 50
Demo adversary
Your single responsibility: adv
Alternatives in Development
- Verdict Reviewer — Adversarial final reviewer for a completed work order or diff — checks the change against its acceptance crite 1.4k ★
- Dashclaw Drift Auditor — Audits DashClaw's drift-prone hardcoded counts and version stamps — SDK method counts (Node/Python), MCP tool/ 297 ★
- Dialogue Reality Enforcer — 对话真实性强制执行器,确保对话符合真实人类交流模式 234 ★
Full documentation available on GitHub
View Source RepositoryRelated Agents
Demo Creator Py
Python creator for the live-demo and workbench builds — the PTY adapter, the terminal websocket route, the det
Demo Orch Stage
Orchestrates the stage phases of the live-demo build — phase-demo-01 (terminal backend), phase-demo-02 (stage
Bugs Edgecases Hunter
Verification Quad Agent 3. Adversarial red-team auditor hunting silent bugs, None/undefined crashes, unhandled
Interleaving Auditor
Concurrency adversary - finds an interleaving where two individually-correct operations, run concurrently, lea
Background Reviewer
Use this agent as the asynchronous deep reviewer that runs off the hot path. Use for routine code review, depe
Docs Auditor
Audits Forge documentation for drift against the code. Checks that README counts, ARCHITECTURE hot paths, SETU
Related Skills
Adversarial Test
Post-build adversarial review — a fresh agent stress-tests recent changes for bugs, regressions, conflicts, an
Adversarial Spec
A Claude Code plugin that iteratively refines product specifications by debating between multiple LLMs until a
Axiom Planning
TDD-validated implementation planning with a four-reviewer plan-review quality gate - atomic-task plans with e