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 instructions/eric-ky-zhang/finprism/agents-mdgit clone --depth 1 https://github.com/Eric-KY-Zhang/FinPrismWrote 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/instructions/eric-ky-zhang/finprism/agents-md)<a href="https://agentmods.dev/instructions/eric-ky-zhang/finprism/agents-md"><img src="https://agentmods.dev/badge/instructions/eric-ky-zhang/finprism/agents-md.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 | $0.01986 | $0.01986 |
| Opus 5 | $0.00993 | $0.00993 |
| Sonnet 5 | $0.00397 | $0.00397 |
| Haiku 4.5 | $0.00199 | $0.00199 |
Grade A, and why
FinPrism 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 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 — 183 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md — FinPrism Agent 协作原则
本文件是 FinPrism 项目的 agent 工作手册（原则层）。FinPrism 是 A/美/港/韩/台 5 市场上市公司财务数据 Excel 桌面工具。
本文档只放原则——可迁移到任何"本地 Excel + VBA + Python 自动化 + 多 agent 协作"型项目。本项目专属的详细规则、命令、文件清单、不变量见
CLAUDE.md。更新同步规则：每次修改 AGENTS.md 必须同步检视 CLAUDE.md 是否需要联动调整（具体规则 / 命令 / 路径常变，原则少变）。
1. Agent 进入项目的第一动作
先读以下文件，不要直接改代码：
CLAUDE.md：本项目专属规则、不变量、回归命令、目录导航。STATUS.md末尾收口段：只从最新阶段读起，再按需回溯。不要默认早期章节仍代表当前状态。ARCHITECTURE.md：做代码改动时必须读，尤其是模块依赖、sheet inventory、数据流、关键 invariants。- 当前 phase plan 或用户最新 prompt：若 plan 与旧文档冲突，以用户最新明确决策为准，但要在报告里说明冲突点。
进入后先判断本轮属于哪类工作，按对应默认策略推进：
| 类型 | 默认策略 |
|---|---|
| Bug 修复 | 先复现，再做最小补丁，再跑相关回归 |
| Phase 实施 | 严格按 plan step 顺序做，不自由重排 |
| UX / 视觉 | 不碰业务逻辑，按截图、cell range、颜色、行高列宽精确执行 |
| 文档收口 | 只更新当前事实，不重写历史决策 |
| Release / 清理 | 保留用户数据，清理可再生数据和敏感缓存 |
2. 三角协作模型
本项目最有效的模式是三角协作：
Planner / Reviewer -> 写 plan、定义边界、独立验收
Generator -> 端到端实现、装表、跑回归、提交报告
Static Auditor -> 扫全量代码，给 backlog，Reviewer 取舍
Agent 要遵守自己的角色边界：
- Generator 不重新设计 Planner 已锁定的方案；遇到 plan 盲点时主动反馈，不擅自扩大范围。
- Reviewer 不只看 Generator 自报，必须独立验证关键路径。
- Static auditor 给建议，不等于必须全做；个人工具项目要按 ROI 砍掉过度工程。
- 多 agent 协作只拆互不冲突的工作面（文档 / inspect / VBA wrapper / Python installer 等），不要让多个 agent 同时改同一 VBA 大模块的重叠区域。
3. 冻结边界：true frozen vs state-bound
这是本项目最重要的边界规则。
| 文件 / 对象类型 | 分类 | 规则 |
|---|---|---|
| 核心业务 helper、公开函数签名、fetch/cache/write/FX 主逻辑 | True frozen | 除非用户明确要求，否则不改 |
集成回归驱动（test_fx_live.py / diff_* 等） |
True frozen | 不为让测试通过而改基线 |
inspect_phase*_state.py |
State-bound | 当主线功能明确删除或移动 sheet/button/cell 时，必须同步 inspect |
STATUS.md / 已闭环 phase 文档 |
历史追溯 | 追加收口段，不改写已闭环历史 |
README.md / ARCHITECTURE.md / CLAUDE.md |
当前说明 | 跟随最终产品状态更新 |
如果 plan 写了 frozen 但用户新需求明确删除被检查对象，要停下或明确说明后再同步 state-bound inspect。不要把过时 inspect 当成产品真相。
4. 数据准确性优先级
财务工具里，错误数字比空值更危险。所有逻辑分支必须遵守：
- 汇率缺失 → 空值 +
FX_MISSING诊断。永不 fallback 到 1。 - 字段缺失 → 空值 + QA / diagnostic，而不是猜测。
- HTTP 失败 → 明确来源、状态、重试次数、缓存状态。
- 跨市场 BS / IS / CF 行项目不可比时，不强行合并；只合并已标准化的 18 项指标。
- QA 规则以低误报为先，WARN 不阻断用户工作流。
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 · 183 lines · 1,986 tokens per session scan A d42b73ecef79
FinPrism AGENTS.md is an instructions file published in the GitHub repository Eric-KY-Zhang/FinPrism (5 stars, last pushed 3mo ago), licensed MIT. It adds 1,986 tokens to every session, about $0.0099 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 instructions, from other repositories
ClawRouter CLAUDE.md
Claude Code instructions for BlockRunAI/ClawRouter, covering clawrouter, commands, project structure, key dependencies and conventions.
InvestSkill GEMINI.md
Gemini CLI instructions for yennanliu/InvestSkill, covering investskill — gemini cli setup & usage guide, installation & setup, quick start, navigate to the investskill directory and start gemini cli (loads gemini.md automatically).
LLM-TradeBot copilot-instructions.md
Copilot instructions for EthanAlgoX/LLM-TradeBot, covering repository instructions, core rules, validation and ai asset governance.
Convertible-Bond-Pricing-Research CLAUDE.md
Instructions for ericxuzhesheng/Convertible-Bond-Pricing-Research, covering claude.md — convertible bond pricing research, current published state and routine commands.
bukio-cli AGENTS.md
AGENTS.md instructions for erikvankempen/bukio-cli, covering agents.md — bukio-cli agent manual, 1. house rules (non-negotiable), 2. environment, 3. command quick reference and 3.1 jurisdiction profiles (16 markets).
swiss-snb-mcp CLAUDE.md
Claude Code instructions for malkreide/swiss-snb-mcp, covering claude.md, teil 1 — konventionen (portfolio-weit), vor der arbeit, tests and wenn etwas rot ist.