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/noshkoto/noshy/agents-mdgit clone --depth 1 https://github.com/Noshkoto/NoshyWrote 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/noshkoto/noshy/agents-md)<a href="https://agentmods.dev/instructions/noshkoto/noshy/agents-md"><img src="https://agentmods.dev/badge/instructions/noshkoto/noshy/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 | $0.00771 | $0.00771 |
| Opus 5 | $0.00385 | $0.00385 |
| Sonnet 5 | $0.00154 | $0.00154 |
| Haiku 4.5 | $0.00077 | $0.00077 |
Grade A, and why
Noshy 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 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.
Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
- **Test with the HTTP API** — `python3 server.py http --port 8721` then curl against `/tools/call`. How it starts
The opening of the file, as written. The whole thing — 62 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Noshy — AI agent instructions
You are working on Noshy, a persistent memory layer for AI agents. It gives agents cross-session memory that accumulates automatically.
What Noshy is
Noshy stores facts, decisions, and preferences extracted from conversations. It injects context at session start so agents don't start from zero. It uses SQLite under the hood, with optional vector embeddings for semantic search, and optional LLM-powered extraction for quality.
Architecture principles
-
Memory should be invisible. The user shouldn't think about it. Extraction happens at session end, injection at session start, everything else is automatic.
-
Dedup aggressively. Storing "fixed the proxy binding" twice is worse than storing it once with higher weight. Jaccard similarity at 40% threshold catches near-duplicates.
-
Search three ways. Keyword for exact finding, semantic for meaning-based recall, graph traversal for connected memories. All three run on every query.
-
Import everything. ICM schema compatibility is a first-class feature. Migration should be one command.
-
Zero deps is a feature. The core runs on Python 3.10 stdlib. Embeddings and LLM extraction are optional layers users opt into.
How to work on this codebase
- Python 3.10+ — no async required, no fancy features. Keep it readable.
- Single-file modules — one concern per file. store.py (data), extractor.py (LLM), embed.py (vectors), context.py (session), server.py (API), hooks.py (automation).
- SQLite with WAL — journal_mode=WAL, busy_timeout=5000. Parameterized queries only. Never string-interpolate user input into SQL.
- Test with the HTTP API —
python3 server.py http --port 8721then curl against/tools/call. - Commit messages — short, active voice. "Fix dedup threshold" not "Fixed dedup threshold".
What not to do
- Don't add new dependencies without a very strong reason. Zero deps is a selling point.
- Don't add a web framework. The stdlib HTTP server is fine for an API that only Hermes talks to.
- Don't over-engineer the schema. ICM compatibility is good. More tables need to prove their worth.
- Don't break the MCP tool contract. If you rename a tool, update both the schema and handler.
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 · 62 lines · 771 tokens per session scan A d06df4ccd5c5
Noshy AGENTS.md is an instructions file published in the GitHub repository Noshkoto/Noshy (5 stars, last pushed 13d ago), licensed Apache-2.0. It adds 771 tokens to every session, about $0.0039 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
hermex AGENTS.md
AGENTS.md instructions for uzairansaruzi/hermex, covering hermex, what makes hermex special?, 1. open at the core, 2. performance without compromise and 3. remote ready.
memohood AGENTS.md
Instructions for maximskorohod/memohood, covering agents.md — установка плагина memohood, что это, способ b — через git (человек ставит вручную, без чата), безопасность and справочные факты (для сверки при установке).
origin-memorycore AGENTS.md
AGENTS.md instructions for moonandecho/origin-memorycore, covering agents.md — origin-memorycore, 项目定位, 开发约定(必须遵守), 常用命令 and 工作流.
hermex CLAUDE.md
Claude Code instructions for uzairansaruzi/hermex, a project described as: Native iPhone app for your Hermes agent.
Hermes-Walkie-Talkie AGENTS.md
AGENTS.md instructions for TheSethRose/Hermes-Walkie-Talkie, covering agents.md, project overview, setup commands, test and validation commands and code style.
remnic AGENTS.md
AGENTS.md instructions for joshuaswarren/remnic, covering remnic - agent guide, architecture boundaries (non-negotiable), upstream references, adapter implementation rules and openclaw compatibility window.