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/trinity-tick/trinity/agents-mdgit clone --depth 1 https://github.com/trinity-tick/trinityWhat 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 | $0.02714 | $0.02714 |
| Opus 5 | $0.01357 | $0.01357 |
| Sonnet 5 | $0.00543 | $0.00543 |
| Haiku 4.5 | $0.00271 | $0.00271 |
Grade A, and why
trinity AGENTS.md scanned grade A with 1 finding 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 yesterday.
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.
Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
curl -s http://127.0.0.1:8001/health How it starts
The opening of the file, as written. The whole thing — 138 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md — Trinity Memory
本文件由 Trinity 生成（2026-08-24 15:26:38）。它让接入本仓库/工作区的 AI Agent 自动了解 Trinity 记忆层的存在、用法与当前状态。 相关规范背景：OpenAI AGENTS.md / Anthropic CLAUDE.md 的"文件即记忆"标准。
Trinity 记忆层实时快照（生成于 2026-08-24 15:26:38）
| 指标 | 值 |
|---|---|
| 会话数 | 240 |
| 结构事件数 | 26663 |
| 目标数 | 71 |
| Todos | 108 |
| 计划 | 0 |
活跃目标（active goals）
| 状态 | 阶段 | 轮次 | 目标 |
|---|---|---|---|
| active | active | 0 | 完成 DeepSeek Harness 迭代升级：全局 @deepseek-ai/dsh 0.1.0-rc.6 → 0.1.0-rc.7(latest)，核对 dsh-trinity 插件版本配套与 ~/.dsh profiles 补丁，重启 web 宿主(:3080)后验证 GUI 与 trinity_* 工具可用， |
| active | active | 0 | 完成旺店通WMS操作规则规范迭代收尾：依据已有9域规则文档与知识来源，补齐并验证三个补充域文件（企业版补充域/帮助手册补充域/跨境版补充域）到 C:/Users/Administrator/KnowledgeBase/AI_WMS/smartcos-wms/docs/，并将迭代结果写入Trinity记忆。 |
| active | active | 0 | 继续执行优化（第三轮）：1) GEN-3 生成侧再优化——KU/SS-P 类目专用提示词 + 上下文剪枝实验（每问只喂最相关 3-5 条 vs 10 条，测噪音对 AnswerAcc 的影响），全量 500q 对比 0.678 基线；2) CH-1 KG/混合通道归因——用 search_hybrid(fusion) |
| active | - | 0 | 按已批准的 Trinity 发展规划执行全部建议：里程碑1（统一 SQuAD 评测口径、修复 Cluster Stress Raft 单 leader 异常、跑官方 LongMemEval-S/LoCoMo/BEAM 1M-10M 基准、同步版本号 v8.2.0）；里程碑2（decay 接入真实 LLM 摘要、决策并落 |
| active | - | 0 | 按 docs_site/optimization-plan.md 的 8 个优化点全方位执行：1) 答案生成评测 harness（mock 500q × DeepSeek → accuracy/latency/cost，逐类目报告）；2) PG FTS GIN 索引 + BEAM 规模复测（对比有/无索引延迟）；3) |
| active | - | 0 | 按优化建议继续执行（第二轮）：1) 生成侧优化——重点提升 TR 时序类目 AnswerAcc（先分析 TR 题结构/失败模式，改进答案提示词与 judge 评分，重跑 500q 对比 0.602 基线）；2) OPT7 后续——修复 Trinity.search 的 mode 参数装饰性问题（semantic/hyb |
| active | - | 0 | 继续执行优化（第三轮）：1) GEN-3 生成侧再优化——KU/SS-P 类目专用提示词 + 上下文剪枝实验（每问只喂最相关 3-5 条 vs 10 条，测噪音对 AnswerAcc 的影响），全量 500q 对比 0.678 基线；2) CH-1 KG/混合通道归因——用 search_hybrid(fusion) |
| active | - | 0 | 执行 EXECUTION_PLAN_V2.md 全部 14 个项目（A1-A5、B1-B5、C1-C4）：每个项目产出可运行工件并验证，修复评测发现的 2 个 API bug，更新执行计划状态，最终汇总完成情况与遗留项。 |
| active | - | 0 | 执行 smartcos-wms 剩余全部建议：①golangci 债务逐包清理（errcheck 249/unused 47/ineffassign 29 等共 328 项，清理后 lint 显著归零或大幅下降且 build/test 全绿）；②实现 12 个 MOCK-ONLY API 端点（carriers/rou |
| active | - | 0 | 按行业操作页形态（扫码优先+任务队列+动作按钮+状态自动流转）重做 SmartCos WMS 全部作业页面：收货作业(ArrivalCounting)、上架作业(Putaway)、称重作业(PackageWeighing)、发货作业(ShipConfirm)、复核作业(VerificationWorkbench)、打包 |
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.
- yesterday First seen · 138 lines · 2,714 tokens per session scan A accd77f6a57b
trinity AGENTS.md is an instructions file published in the GitHub repository trinity-tick/trinity (1 stars, last pushed 3d ago), licensed MIT. It adds 2,714 tokens to every session, about $0.0136 per session on Opus 5. A static security scan graded it A with 1 finding (makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other instructions, from other repositories
engram-mcp CLAUDE.md
Instructions for edg-l/engram-mcp, covering engram mcp, development rules, structure, key types and mcp capabilities.
pensyve copilot-instructions.md
Copilot instructions for major7apps/pensyve, covering pensyve working-memory substrate for vs code copilot, memory reflex rule, "substantive answer" means, "lesson lands" means and memory classification.
pensyve AGENTS.md
AGENTS.md instructions for major7apps/pensyve, covering pensyve — agents.md, what this repo is, start here, directory map and build & dev commands.
pensyve CLAUDE.md
Claude Code instructions for major7apps/pensyve, covering claude.md, claude code specifics and see also.
mnemic CLAUDE.md
Claude Code instructions for dongtang3/mnemic: This project includes a project-level .mcp.json for the Mnemic memory MCP server.
mnemic AGENTS.md
AGENTS.md instructions for dongtang3/mnemic: Mnemic is a graph-backed long-term memory substrate for coding agents and LLM applications.