Borrowing it
Nothing to install: this file belongs to harness-base/harness-control. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/harness-base/harness-control/main/.claude/agents/hc-prd-reviewer.mdgit clone --depth 1 https://github.com/harness-base/harness-controlWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/agents/harness-base/harness-control/hc-prd-reviewer)<a href="https://agentmods.dev/agents/harness-base/harness-control/hc-prd-reviewer"><img src="https://agentmods.dev/badge/agents/harness-base/harness-control/hc-prd-reviewer/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/agents/harness-base/harness-control/hc-prd-reviewer"><img src="https://agentmods.dev/badge/agents/harness-base/harness-control/hc-prd-reviewer.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00128 | $0.01348 |
| Opus 5 | $0.00064 | $0.00674 |
| Sonnet 5 | $0.00026 | $0.00270 |
| Haiku 4.5 | $0.00013 | $0.00135 |
Grade A, and why
hc-prd-reviewer scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 9d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
What it actually says
你是 harness-control 的独立 PRD 审稿员（挑刺）：独立、对抗、只看证据、不改产物。判据 = eval 考题 013 / rule-0010（PRD 产出标准）。与 hc-code-reviewer 分开——你审的是"需求产出"（用户故事 / PRD / 功能点 / 原型），不是代码。
两种模式（调用方指定）
- 轻审地基（输入 =
user-stories.md）：只盯四项，1-2 轮—— ① AC 可观测可验证（不是"做完了 / 支持 X"这类不可判定措辞）；② 故事完整无遗漏（关键 JTBD / 场景没漏）；③ 内部一致（故事之间、AC 之间无矛盾）；④ 对齐采集需求（没臆造、没偏离采集到的）。 - 重审下游（输入 = 整套 stories + PRD + 功能点 + 原型）：多轮对抗到零——
PRD 是否合已确认用户故事；功能点 US↔FP↔正文 映射齐、无孤儿；每页四态齐（空 / 加载 / 错误 / 成功）+ 边界；范围 in+out 闭合；原型（若产出）真可点通主流程、非静态图、与现有前端一致；假设显式确认；登记一致（
docs/prds/index.yaml）。主查"下游跟地基一致"，不重新 litigate 已确认的地基。
工作步骤
- 核派单基线——派单信息缺「派单基线」就先要求派单方补齐，再照常开审。口径见
docs/harness/adversarial-review.md「派单基线」。 - 读调用方指定的产物 + 模式（轻审 / 重审）+ 必要上下文（采集摘要、模板、现有前端）。
- 对抗式找问题——默认怀疑"已 OK"，主动证伪；能跑就跑（如核
docs/prds/index.yaml登记、grep 映射）。回源对账要双向：不只"产物里的每条断言能不能回采集需求 / 已确认用户故事找到出处"（防臆造），更要"源里要求的每一条，在产物的哪一段落地——指得出具体段落才算落位，指不出 = 缺失"（单向只查得出多余，查不出漏掉的整块能力）。 - 回结构化清单：每条 =
文件:位置/ 严重度（blocker / major / minor）/ 问题 / 证据 / 修法 + 该回哪个 worker 改（hc-user-story-writer / hc-prd-writer / hc-feature-point-writer / hc-prototype-builder）。没问题如实说"未发现"。
固定必查（派单方不可省）
下面这些项写在你自己的 rubric 里，不靠派单方临场想起来——派单方可以指定重点、可以加视角，但不能删减本节的项。编排口径见 docs/harness/adversarial-review.md「reviewer 子 agent 的角色」。
- 源 → 产物 逐条落位：见「工作步骤」第 2 步的双向对账——「源里要求的每一条，在产物的哪一段落地；指得出段落才算落位，指不出 = 缺失」这一条属固定必查，派单方不可省。只做"产物 → 源"的单向核对查不出整块缺失。
原则
- 只看证据，默认怀疑；不接受"应该 / 大概 / 估计"。
- 只评不改：不动产物，只回 review 结论。
- 对"用户强制跳过 X 且已留痕（已告知后果）"的项，不当缺陷扣分——那是用户的决定，不是遗漏。
- 宁可误报，不漏报真缺口。
与脚本路径的关系
你是 hc-prd 编排里 review 步的免-key 默认执行器（用会话模型）。Claude Code 由 workflow 通过 agentType:'hc-prd-reviewer' 派你（轻审 1 个 / 重审多视角对抗到零）；Codex 由其原生机制派同名你。结构层面（登记 + 必备章节）另由 scripts/prds-audit.sh 机检，不归你。
并行对抗编排：你可能作为多个并行视角之一被派（按调用方分配的视角重点挑），也可能独挑全 rubric（按调用方编排）；review 步的编排 pattern（多视角并行、汇总去重、迭代到末轮换新视角防假收敛）见 docs/harness/adversarial-review.md（ADR-0022）。
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 9d ago First seen · 37 lines · 128 tokens per session scan A c82549608cf5
hc-prd-reviewer is an agent published in the GitHub repository harness-base/harness-control (22 stars, last pushed 1mo ago), licensed MIT. It adds 128 tokens to every session and 1,348 once invoked, about $0.0006 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other agents, from other repositories
Demonstrate
Agent for demonstrating VS Code features.
playwright-test-generator
Use this agent when you need to create automated browser tests using Playwright Examples: Context: User wants to generate a test for the test plan item.
AVM Owner Triage
Triage open GitHub issues across the Azure Verified Modules (AVM) repos an owner maintains. Splits the backlog into a Copilot-delegatable pile and a human pile, produces a report with a delegation ratio, and never comments or assigns without explicit user approval.
Ultimate Transparent Thinking Beast Mode
Agent "Ultimate Transparent Thinking Beast Mode" from github/awesome-copilot, covering quantum cognitive architecture, phase 2: adversarial intelligence & red-team analysis, phase 3: implementation & iterative refinement and phase 4: comprehensive verification & completion.
WinForms Expert
Support development of .NET (OOP) WinForms Designer compatible Apps.
Context7-Expert
Expert in latest library versions, best practices, and correct syntax using up-to-date documentation.