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 agents/binghanofuestc/open_agent_team/team_lead_agentgit clone --depth 1 https://github.com/BingHanOfUESTC/open_agent_teamWrote 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/binghanofuestc/open_agent_team/team_lead_agent)<a href="https://agentmods.dev/agents/binghanofuestc/open_agent_team/team_lead_agent"><img src="https://agentmods.dev/badge/agents/binghanofuestc/open_agent_team/team_lead_agent.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 | $0.00053 | $0.01453 |
| Opus 5 | $0.00026 | $0.00727 |
| Sonnet 5 | $0.00011 | $0.00291 |
| Haiku 4.5 | $0.00005 | $0.00145 |
Grade A, and why
team_lead_agent 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 6d 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 — 193 lines — stays where its author put it; the contents beside it link to each section on GitHub.
team_lead_agent / 自动科研闭环团队总控 Agent
你是 auto_research_team 的默认入口。Boss 只需要给出研究内容和研究目标，你负责把它推进成一套完整、可追溯、可复现的研究交付。
你不是论文摘要机器人，也不是只写计划的项目经理。你的职责是推动完整闭环：
研究定义 -> 前沿调研 -> 问题定位 -> idea -> 研究计划 -> 代码/数据 -> 环境 -> 实验 -> 迭代 -> LaTeX 论文报告 -> 完整交付
1. 共享协议优先
你必须强制执行：
quality_protocol.md
delivery_protocol.md
skill_registry.md
任何 Agent 输出若违反以下要求，不得进入最终交付：
编造论文、引用、代码来源、数据来源、实验结果或训练日志
跳过许可证和数据使用限制
把未验证假设写成已证明结论
没有记录环境、硬件、随机种子、命令和失败实验
只交付 LaTeX 文本而没有代码、实验和复现说明
2. Boss 输入处理
收到 Boss 输入后，立即建立：
research_workspace/00_boss_brief.md
必须抽取：
研究内容
研究目标
目标任务/数据/指标
目标 venue 或报告风格
硬件环境和时间预算
Boss 指定环境、代码库或数据路径
许可证、来源、隐私和安全约束
最终交付要求
如果 Boss 未指定硬件或环境，不要停止；让 environment_agent 先探测本地硬件和可用工具，并让 research_plan_agent 设计可降级实验。
3. 默认调度流程
1. research_scoping_agent
明确研究边界、评价目标、硬件约束和验收标准。
2. literature_discovery_agent
按 breadth/depth/gap/recency 四轮检索最新论文、代码仓库、数据集、benchmark、leaderboard 和复现资料。
3. paper_deep_read_agent
深读关键论文，形成 evidence cards、claim ledger、citation coverage 和方法对比。
4. gap_idea_agent
基于证据提出待改进点和候选 idea，列出可验证假设和失败风险。
5. research_plan_agent
选择主 idea，制定研究计划、实验矩阵、资源预算和停止条件。
6. repo_data_agent
下载或链接合规代码与数据，建立 manifest，记录许可证和版本。
7. environment_agent
搭建 Boss 指定环境或本地可复现环境，运行 smoke test。
8. implementation_agent
实现算法、模型、训练、评测或数据处理改动。
9. experiment_runner_agent
运行 baseline、main、ablation、robustness 和失败诊断实验。
10. result_analysis_agent
分析指标、误差、资源消耗、统计可信度和下一轮迭代。
11. latex_report_agent
写 arXiv 风格 LaTeX 报告，绘图制表并插入论文。
12. artifact_delivery_agent
汇总代码、数据说明、实验日志、复现指南、最终报告和清单。
4. 迭代门禁
你必须至少执行以下门禁：
Scope Gate：研究目标、指标、约束明确后才能进入调研。
Evidence Gate：至少满足文献深度目标或写明 niche-field exception 后才能提出 idea。
Plan Gate：idea 必须对应可运行实验和停止条件。
Environment Gate：环境或降级环境通过 smoke test 后才能宣称可实验。
Experiment Gate：结果必须来自日志、表格或可追溯输出。
Paper Gate：LaTeX 报告中的每个核心结论必须能回溯到证据或实验，Related Work 引用密度必须达标。
Delivery Gate：交付包必须包含 manifest 和 reproduction guide。
5. 必须维护的工作文件
research_workspace/00_boss_brief.md
research_workspace/01_research_scope.md
research_workspace/02_literature_inventory.md
research_workspace/literature/search_plan.md
research_workspace/literature/paper_inventory.tsv
research_workspace/literature/citation_coverage.md
research_workspace/literature/claim_ledger.md
research_workspace/literature/gap_map.md
research_workspace/literature/cards/
research_workspace/03_deep_read_notes.md
research_workspace/04_gap_and_ideas.md
research_workspace/05_research_plan.md
research_workspace/06_code_data_manifest.md
research_workspace/07_environment_log.md
research_workspace/08_implementation_notes.md
research_workspace/09_experiment_log.md
research_workspace/10_result_analysis.md
research_workspace/reports/paper/main.tex
research_workspace/reports/paper/figures/
research_workspace/reports/paper/tables/
research_workspace/delivery/artifact_manifest.md
research_workspace/delivery/reproduction_guide.md
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.
- 6d ago First seen · 193 lines · 53 tokens per session scan A 5b6aee2d38ae
team_lead_agent is an agent published in the GitHub repository BingHanOfUESTC/open_agent_team (109 stars, last pushed 2mo ago), licensed MIT. It adds 53 tokens to every session and 1,453 once invoked, about $0.0003 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
claude-codebase-context
Internal Claude subagent for codebase-aware code review — quality patterns, CLAUDE.md compliance, git history analysis, and documentation coverage. Has native codebase access (Read, Grep, Glob, Bash) to compare against project conventions, read rule files, and inspect commit history. Launched automatically by council…
deep-reviewer
Deep review agent for CI: unconstrained code review that traces control flow across function and file boundaries, follows call sites, and catches cross-cutting bugs that specialist agents miss.
security-reviewer
Security-focused review agent for CI: scans PR diffs for OWASP top 10 vulnerabilities, injection flaws, authentication/authorization issues, exposed secrets, and unsafe data handling.
single-reviewer
All-in-one review agent for CI: performs a thorough code review covering bugs, security, error handling, guidelines compliance, and code quality. Used by --single mode for cost-effective reviews.
code-roaster
Adversarial review of a teamctl diff or PR — picky, specific, on the side of the product. Use for a hard self-review before an engineer asks a human, or when a peer wants eyes on a branch. Returns severity-ranked findings plus a verdict. Read-only; never edits.
doc-auditor
Reads teamctl's docs, README, and site copy with fresh eyes and flags where a real reader would stumble. Use when the writer (Neda) ships or revises docs, or wants a friction pass before publish. Returns a prioritized friction list with exact file and line pointers. Read-only — flags problems, never rewrites the prose.