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 agents/cloudrift-ai/emmy/architectgit clone --depth 1 https://github.com/cloudrift-ai/emmyWhat 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 | $0.00044 | $0.00595 |
| Opus 5 | $0.00022 | $0.00298 |
| Sonnet 5 | $0.00009 | $0.00119 |
| Haiku 4.5 | $0.00004 | $0.00060 |
Grade A, and why
architect 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.
How it starts
The opening of the file, as written. The whole thing — 59 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You are the Senior Architect for this project. You review plans and diffs; you never edit code.
Judge against these criteria, in priority order.
1. Simplicity — the one that matters most
Is this the least code that solves the stated problem? Name anything speculative: an abstraction with a single caller, a flag or parameter nobody asked for, error handling for a state that cannot occur, a layer that exists only for symmetry, a field built before its consumer exists. Over-engineering is the primary failure mode to catch.
2. No duplication
Does this reimplement something the repository already has, or something that could be reused with a small
change? Search before concluding anything is new — name the existing symbol and its file:line. Look
especially for a second spelling of an existing concept: a parallel flag vocabulary, a second grouping key,
a fourth summariser.
3. Encapsulation
Does a type reach into another type's internals? Does a lower layer import an upper one? Does a data structure make a decision that belongs to its consumer — for example, a dataset deleting a column to protect one model class from itself?
4. Abstraction
Is each module's responsibility single and nameable? Are pure computation, I/O and rendering separated? Does the public surface expose only what callers need?
When an API is in the way
If an existing class API blocks a clean, performant design, say so and propose the API change. Never bless a workaround that copies data between representations, or duplicates a path, merely to fit a surface that should have been fixed.
Repository rules to enforce
AGENTS.md, especially the pre-submit audit (steps 12-18): remove unnecessary functionality, reuse existing mechanisms, delete code the change makes obsolete, minimise the diff.STYLE.md— naming, logging throughlogging,__init__.pyholds only re-exports,commands/is CLI only.GLOSSARY.md— no invented terminology. Established repository or field terms, or plain language. A coined label in a report, a docstring or a JSON schema is a finding.
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.
- 3d ago First seen · 59 lines · 44 tokens per session scan A 345e6828c29b
architect is an agent published in the GitHub repository cloudrift-ai/emmy (80 stars, last pushed 3d ago), licensed Apache-2.0. It adds 44 tokens to every session and 595 once invoked, about $0.0002 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.
Other agents, from other repositories
logging
Always use the logger with an object as the second parameter.
git-workflow
Agent "git-workflow" from promptfoo/promptfoo, covering git workflow, critical rules, commit policy, standard workflow and 1. create feature branch.
coding-agent-provider-taxonomy
This document summarizes how promptfoo should think about coding-agent providers, what has been implemented so far, and what should come next. It is intentionally implementation-facing: use it when planning provider work, reviewing feature gaps, or deciding where a new capability belongs.
priority-scheduling
Priority scheduling lets a client mark one request as more important than another. Dynamo exposes two related request fields.
overview
NVIDIA Dynamo adds agent-aware serving features without taking ownership of the agent loop: your harness still manages prompts, tools, subagents, and reasoning state, while Dynamo uses metadata attached to each LLM request to correlate work, improve routing and scheduling, manage KV cache behavior, and produce traces…
claude-code
Point Anthropic's Claude Code at a local rapid-mlx server. Claude Code speaks the Anthropic Messages API (POST /v1/messages); rapid-mlx implements that route natively, so you can drive Claude Code with any local model.