Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add commands/brightbear2026/research-agent/deep-researchgit clone --depth 1 https://github.com/brightbear2026/research-agentWrote 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/commands/brightbear2026/research-agent/deep-research)<a href="https://agentmods.dev/commands/brightbear2026/research-agent/deep-research"><img src="https://agentmods.dev/badge/commands/brightbear2026/research-agent/deep-research.svg" alt="Measured on agentmods" 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 | $0.00024 | $0.04070 |
| Opus 5 | $0.00012 | $0.02035 |
| Sonnet 5 | $0.00005 | $0.00814 |
| Haiku 4.5 | $0.00002 | $0.00407 |
Grade A, and why
deep-research scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 5d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 85 lines — stays where its author put it; the contents beside it link to each section on GitHub.
你现在是「深度研究」流程的调度方（orchestrator）。课题为：$ARGUMENTS（若为空，先用 AskUserQuestion 向用户索要课题、背景、目标读者、研究范围与重点关注问题）。
严格遵循本仓库 CLAUDE.md 的全部约定。按下面 6 阶段推进；确认点和停止位置必须服从 tools/workflow_policy.py 的状态结果，不得只凭提示词自行判断。
0. 解析与初始化
- 每个课题独立启动，不前置参考
projects/下其他课题的历史报告/产出。新课题从课题本身的研究问题与关键词出发；既有课题报告仅作交付归档，不作为新课题的背景输入。除非用户明确要求引用某既有课题结论，否则不读取其他课题目录。 - 从参数解析
depth（默认「标准」）。规则：- 快速：3–5 个核心论证章节，正文约 1–2 万字；重要结论 ≥1 个 A/B 级来源，数值声明仍须 ≥2 个独立来源；截图按需。
- 标准：5–8 个论证章节，正文约 2–4 万字；重要结论 ≥2 个独立来源且至少含 A/B 级；关键页强制截图。
- 深度：6–10 个论证章节，正文原则上不超过 6 万字，标准要求 + 每章反向验证多轮 + 全量截图。
- 从参数解析
mode（默认regular）：regular=阶段一/二/三各确认一次；plan=仅大纲后集中确认一次并停止；execution=仓库流程零确认。三种模式都可访问公开来源，但不得绕过登录、验证码、付费墙或网站访问控制。 - 项目目录名 =
projects/<课题slug>（如「AI Agent 安全」→projects/ai-agent-security；slug 用课题英文/拼音短名、kebab-case）。每个课题一个独立 slug 文件夹，绝不复用旧目录，从源头避免覆盖上一课题。 - 运行
uv run python tools/scaffold.py projects/<课题slug>建骨架。若该目录已存在且非空，scaffold 会拒绝（需换 slug；或加--force——会先把旧目录备份为<slug>-backup-<时间>，不直接清空）。 - 运行
uv run python tools/workflow_policy.py init --root projects/<课题slug> --mode <mode>初始化机器可执行状态。每阶段完成后运行advance：退出码 3 才询问用户，确认后用advance --confirmed；退出码 4 表示章节未登记或尚未全部完成，必须留在阶段四；返回stop必须停止。外部来源失败用record-failure登记，重试预算耗尽后转资料缺口继续，不得无限循环。章节级重派受config/workflow_modes.yaml的max_chapter_attempts（默认 3）约束：达上限后chapter --status in_progress会被拒（返回非 0 + 升级提示），须升级检索策略、换维度或转资料缺口，不重派——这是对研究型子代理空转的兜底熔断。 - 预检当前宿主能否发现
diagram-designskill。可用时在课题根写.diagram-design，内容为用户显式/已保存 profile；没有指定时写profile: default。不可用时记录“Mermaid fallback”，不得因此阻断研究或新增确认点。该插件只生成解释性图，不改变证据标准。 - 用
TaskCreate建立阶段一~六的任务，逐个in_progress/completed。
阶段一 · 研究启动（先调研再设计大纲）
产出（写入 sources/stage1_kickoff.md）：
- 课题定义与边界（研究对象、不研究什么、时间/地域/分析层级范围）。
- 概念分解与辨析（关键，不可跳）：若课题标题含多个概念（如「A 与 B」），必须把每个概念作为对等概念独立定义，并辨析它们的关系（包含/交叉/因果/层级）+ 一张概念关系图。这些定义须作为 Ch1「§概念定义与辨析」节写进报告正文（不只停在过程文档）——这是最常被漏掉的环节。
- 术语表
data/glossary.md（用 Write 创建）：列出全部核心术语（中英、定义、与相邻概念的区别），作为全报告统一口径；Phase 4 每个 researcher 须先读。 - ≥15 个研究问题，按类划分（定义/历史/现状/数据/技术/企业/人物观点/竞争/应用/风险/趋势/战略）。
- 中英文关键词矩阵（中/英/学术/企业/政策/数据/券商与投行研报/反方/争议/年份/地域；按需加入“行业深度/公司深度/盈利预测/一致预期/equity research/sell-side research/industry report”等组合词）。
- 搜索策略与预期资料类型、研究风险。
→ 完成后调用状态机
advance；仅当返回needs_confirmation时用AskUserQuestion确认，再advance --confirmed。
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.
- 5d ago First seen · 85 lines · 24 tokens per session scan A 31b7f371edc9
deep-research is a command published in the GitHub repository brightbear2026/research-agent (2 stars, last pushed 16d ago), licensed MIT. It adds 24 tokens to every session and 4,070 once invoked, about $0.0001 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 2026-08-31.
Other commands, from other repositories
checklist
Generate a custom checklist for the current feature based on user requirements.
clarify
Identify underspecified areas in the current feature spec by asking up to 5 highly targeted clarification questions and encoding answers back into the spec.
specify
Create or update the feature specification from a natural language feature description.
analyze
Perform a non-destructive cross-artifact consistency and quality analysis across spec.md, plan.md, and tasks.md after task generation.
converge
Assess the current codebase against the feature's spec, plan, and tasks, then append any remaining unbuilt work as new tasks to tasks.md so implement can complete it.
implement
Execute the implementation plan by processing and executing all tasks defined in tasks.md.