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/amitnextech/trailmem/agents-mdgit clone --depth 1 https://github.com/amitnexTech/trailmemWhat 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 | $0.01648 | $0.01648 |
| Opus 5 | $0.00824 | $0.00824 |
| Sonnet 5 | $0.00330 | $0.00330 |
| Haiku 4.5 | $0.00165 | $0.00165 |
Grade A, and why
trailmem 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 2d 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 — 76 lines — stays where its author put it; the contents beside it link to each section on GitHub.
trailmem — Agent Guide
trailmem is a local-first, graph-linked persistent-memory MCP server for AI coding agents (Python, open-source, MIT). This repo is in the design → implementation transition: the full spec is locked (Q1–Q16); code is being written now.
This file is the single source of truth for any coding agent working here. CLAUDE.md just imports it (@AGENTS.md); other agents (Codex, Kiro, Kilo, OpenCode, Cursor, Gemini) read this file directly.
Session boot sequence — do this first
- Read the auto-loaded build-log memory (
MEMORY.mdfor Claude; the trailmem MCP briefing block if a pinned/briefing is already injected into context). Do NOT calltrailmem_welcomeif a briefing is already present — it just wastes tokens. - Read the relevant
docs/page for the task (via the open-knowledge MCP, see below) before touching code. - Before non-trivial work, query trailmem for prior decisions/lessons instead of re-deriving state or re-asking:
trailmem_query("<topic>"). Check the pinned constraints first — they carry standing rules. - Only then start work. Not doing steps 1–3 is a known cause of re-deriving locked design and token burn.
Tool usage rules — pick the right tool for the question
| Question type | Use this | NOT this |
|---|---|---|
| Codebase / file relationships / "how does X work" | graphify (graphify_query_graph, graphify_shortest_path, graphify_get_node) — graph lives at graphify-out/graph.json |
raw grep/Glob/source reading from scratch |
Design spec / docs/*.md |
open-knowledge MCP (mcp__open-knowledge__search, mcp__open-knowledge__exec) — start at docs/index.md |
native Read/Grep/Glob on docs/*.md |
| Past decisions / lessons / cross-session context | trailmem MCP (trailmem_query, trailmem_show) |
re-asking the user, or guessing |
| Storing a decision/lesson/task | trailmem MCP trailmem_store (English, project-scoped path /home/amit/trailmem, link every memory) |
only file-notes that never reach trailmem |
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.
- 2d ago First seen · 76 lines · 1,648 tokens per session scan A f90341c10d8e
trailmem AGENTS.md is an instructions file published in the GitHub repository amitnexTech/trailmem (1 stars, last pushed 1mo ago), licensed MIT. It adds 1,648 tokens to every session, about $0.0082 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
palace-rs AGENTS.md
Instructions for AncientiCe/palace-rs, covering agent rules, 1. test-driven development (tdd), 2. quality gates on every task, 3. no plan markdown files and 4. internal api discipline.
Open-Neural-Substrate CLAUDE.md
Claude Code instructions for pedroknigge/Open-Neural-Substrate, covering claude.md — instructions for claude code, what this project is, architecture summary, key design rules and tech stack.
palace-rs CLAUDE.md
Instructions for AncientiCe/palace-rs, covering claude notes, rtk (rust token killer) - token-optimized commands, golden rule, ❌ wrong and ✅ correct.
remnic AGENTS.md
Instructions for joshuaswarren/remnic, covering remnic - agent guide, architecture boundaries (non-negotiable), upstream references, adapter implementation rules and openclaw compatibility window.
engraphis AGENTS.md
Instructions for Coding-Dev-Tools/engraphis, covering agents.md — engraphis, 0. read this first — two architectures live in one package, 1. commands, ── unified dashboard + memory inspector ── and 2. the v2 recall pipeline (where the real work is).
gno AGENTS.md
Instructions for gmickel/gno, covering knowledge cutoff warning, apis - bun first!, must use bun, acceptable node:\ (no bun equivalent) and testing.