agent-pulse AGENTS.md

Project instructions for Agent Pulse, a Chinese-first information system that turns AI industry events, research, funding, policy, and online signals into evidence-backed findings. It distinguishes reported facts from interpretations and predictions.

In plain words
What is it for?
Working on event collection, evidence tracking, topic and company relationships, trend analysis, ranking, publishing, and research workflows.
Why use it?
They set rules for trustworthy research, including source evidence, uncertainty labels, and separation of facts from opinions. This helps prevent unsupported claims and misleading reports.

Instructions file for CodexOpenCode

Install

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.

agentmods
npx agentmods add instructions/barretlee/agent-pulse/agents-md
Clone the repo
git clone --depth 1 https://github.com/barretlee/agent-pulse

Made for: Codex, OpenCode.

Per session 2,533 This file is loaded in full into every session.
When invoked 2,533 The same file — it is already loaded in full.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

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.

ModelPer sessionOnce invoked
Fable 5 $0.02533 $0.02533
Opus 5 $0.01267 $0.01267
Sonnet 5 $0.00507 $0.00507
Haiku 4.5 $0.00253 $0.00253

Measured 3d ago against content hash 80fb8ab41512, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

agent-pulse AGENTS.md 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 3d 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.

AGENTS.md · 102 lines

How it starts

The opening of the file, as written. The whole thing — 102 lines — stays where its author put it; the contents beside it link to each section on GitHub.

AGENTS.md

项目使命

Agent Pulse 不是新闻聚合器,而是一个中国优先、全球覆盖的 AI 行业认知与行动系统。它面向 CEO、投资负责人、创业者、业务负责人和技术负责人,把发布、论文、资本动作、政策与跨平台传播信号收敛为可验证、可复述、可行动的行业主线。

每个公开事件都应回答:发生了什么、证据是什么、为什么重要、影响谁、下一步观察什么、可以采取什么动作。

协作约定

  • 产品文案、PRD、规格、运营说明默认使用中文;代码、标识符、注释和提交信息默认使用英文。
  • 修改架构、schema、采集、排名、发布、来源生命周期或星探能力前,先阅读 docs/specs/ 中对应规格;跨层变更必须 docs-first。
  • 搜索现有实现后再编辑,优先扩展既有模块,保持 diff 小、可回滚。
  • SQLite 是零配置默认数据库;只有通过 MySQL 集成验证后,才能宣称某项能力兼容 MySQL。
  • 使用 npm;提交前执行 npm run check,并补充与风险相称的契约、集成和浏览器测试。
  • 不得把 seed/demo、预留字段、设计文档或未被前台消费的配置描述成已实现能力。

情报模型与主线

  • Event 是唯一事实节点;TrackActorAudienceRegionTechnologyResourceOpportunity 只附着解释,不复制事实。
  • 必须同时维护模型能力与研究、Agent 与软件重构、产品与商业验证、基础设施与成本、资本与公司演化、全球创新版图,以及 To C / To B / To D / To G 支线。
  • 支线至少能按技术、地域、公司、人物、政策、场景和时间阶段组织,并与主线共享证据。
  • 主线不是普通筛选标签:应逐步支持阶段、里程碑、转折、因果、对比和阶段性总结。
  • “角色已收录”与“角色已被有效观测”必须分开表达;跟随、并跑、领先、受限或出海判断必须绑定证据节点。

事实、判断与热点

  • 明确区分事实、推断、观点、预测和机会假设;不确定内容必须标注证据状态与置信度。
  • 聚合站只能用于候选发现、聚类提示或传播热度,不能作为重大事实的唯一证据。
  • 公开事实原则上至少需要一个 Tier 1 原始证据,或两个相互独立的 Tier 2 证据;例外必须明确标记为“待证实”。
  • 每个公开事件必须回链可点击的原始证据;禁止发布第三方完整正文。
  • 热点必须来自真实传播证据,不得把手工分数包装成已测量热度。
  • 来源独立性按 source、author 和 media group 判断,不能用来源角色或权威分代替身份去重。
  • 热度至少考虑独立作者、独立来源、平台宽度、地区/语言宽度、扩散速度、持续性、转载矩阵和时间衰减。
  • 人工覆盖评分必须保留原值、原因、时间和审计记录。

数据源是一等资产

  • 所有 collector 必须位于统一 SourceAdapter 契约之后;编排层不得出现来源特例。
  • Source 至少记录 owner、tier、role、region、language、acquisition、license/robots、frequency、quota、authority、freshness SLO、adapter version、health 和 lifecycle state。
  • 生命周期统一按 draft -> shadow -> active -> degraded -> quarantined -> retired 演进;shadow 是隔离验证阶段,retired 是保留历史 provenance 的软卸载。
  • 新来源默认 draftshadow;只有通过契约测试、合规检查和真实运行窗口后才能晋级 active
  • adapter 应具备 capability manifest、配置 schema、规范化 schema、cursor/cache state、fixture、contract test、health probe 和版本迁移/卸载策略。
  • 来源管理应支持发现、试拉取、启用、停用、降级、隔离、恢复和软卸载;退役不得破坏历史 provenance。
  • 高价值来源地图必须区分已接入、待接入、受限、替代来源和覆盖缺口,并显示国内/海外、官方/专家/社交覆盖率。
  • 博主、学者、CXO、记者和投资机构必须维护身份、机构关系和媒体矩阵归属,避免制造虚假独立来源数。

采集稳定性与安全

  • 优先使用官方 API、RSS、Atom、GitHub Releases 和稳定公开数据;遵守 robots、许可和速率限制,不绕过登录、WAF、CAPTCHA 或付费墙。
  • 请求必须具备 timeout、有限 retry、指数 backoff + jitter、Retry-After、并发/速率预算和流式响应体上限。
  • 每个 HTTP redirect 都必须重新执行 SSRF 检查;禁止 loopback、link-local、私网、凭据 URL 和 DNS rebinding 风险。
  • ETag、Last-Modified、cursor 和 fingerprint 必须原子持久化;失败不得推进 cursor。
  • 单来源失败不得中断整批任务;连续失败应触发 circuit breaker 和降级,恢复需经过 probe/shadow verification。
  • 每次运行记录 source、adapter version、耗时、状态、item counts、cursor、错误分类和重试次数;parser 漂移、异常空结果、数量突变和时间异常应可见。
  • 新增或修改来源时,必须同时提供 fixture、成功路径、失败/降级路径和 schema 漂移测试。

Read the full file on GitHub · 102 lines

Changes

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.

  1. 3d ago First seen · 102 lines · 2,533 tokens per session scan A 80fb8ab41512

Subscribe to this mod's changes

agent-pulse AGENTS.md is an instructions file published in the GitHub repository barretlee/agent-pulse (236 stars, last pushed 6d ago), licensed MIT. It adds 2,533 tokens to every session, about $0.0127 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.