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/diegosouzapw/omniglyph/agents-mdgit clone --depth 1 https://github.com/diegosouzapw/OmniGlyphWhat 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 | $0.00854 | $0.00854 |
| Opus 5 | $0.00427 | $0.00427 |
| Sonnet 5 | $0.00171 | $0.00171 |
| Haiku 4.5 | $0.00085 | $0.00085 |
Grade A, and why
OmniGlyph 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 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.
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 — 72 lines — stays where its author put it; the contents beside it link to each section on GitHub.
omniglyph — Agent Guidelines
Project
Context-as-image compression proxy for LLMs. Rewrites the bulky parts of each
request (system prompt, tool docs, old history, large tool outputs) into dense
1-bit PNG pages before they leave the machine, using exact per-provider
billing math so conversion happens only when it is profitable. Ships as a
standalone proxy (Node + Cloudflare Workers) and as the omniglyph compression
engine inside OmniRoute.
Stack
TypeScript (strict, ESM) · Vitest · esbuild (scripts/build.mjs) · tsx ·
Cloudflare Wrangler · pnpm. Node ≥18.
Build / Test Commands
pnpm install && pnpm run build # bundle to dist/
pnpm run typecheck # tsc --noEmit — must be clean
pnpm test # vitest run (full suite; slow — background it)
npx vitest run tests/<file>.test.ts # focused run while iterating
Architecture
src/core/transform.ts— the pipeline: pick bulky blocks, gate on profitability + model approval, reflow text, render, splice pages back into the request (cache-friendly ordering).src/core/anthropic-vision.ts— exact Anthropic image billing: 28px patches- per-block overhead, per-tier resize caps, page geometry. Both tiers render the standard 1568×728 page (the high-res tier is a measured billing trap).
src/core/gemini-model-profiles.ts— Gemini billing formulas (tiles +media_resolution) and tile-native page geometry.src/core/render*.ts+assets/— 1-bit 5×8 glyph atlas rendering to PNG; deterministic renders are memoized (render-cache.ts, LRU).src/core/proxy.ts— request handling, refusal-retry sniffer (SSE/JSON replay withretryRefusalWithOriginalkill switch), savings accounting against acount_tokenscounterfactual.src/node.ts— local host: dashboard, events file, upstream routing.src/worker.ts— Cloudflare Workers host.src/core/index.ts— public package API; billing math (anthropicImageTokens,resolveAnthropicVisionTier, tier caps) is re-exported at the root because OmniRoute's image-aware token estimator consumes it.benchmarks/billing-sweep/— proves the billing formulas (residual zero).benchmarks/density-frontier/— read-accuracy frontier per model/density; transports: direct API, OpenRouter,--via-cli,--via-omniroute.
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 · 72 lines · 854 tokens per session scan A 58759558fd4e
OmniGlyph AGENTS.md is an instructions file published in the GitHub repository diegosouzapw/OmniGlyph (104 stars, last pushed 7d ago), licensed MIT. It adds 854 tokens to every session, about $0.0043 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 instructions, from other repositories
tokenfold AGENTS.md
Instructions for snchimata/tokenfold, covering repository guidelines, working principles, repository map, implementation conventions and validation.
distil CLAUDE.md
Claude Code instructions for munhq/distil, covering distil — context optimization for llm agents, what this is, consumption modes, architecture and layers (src/layers/).
densely CLAUDE.md
Instructions for alibaizhanov/densely: Behavioral guidelines to reduce common LLM coding mistakes. Merge with project-specific instructions as needed.
DeepSpeed AGENTS.md
AGENTS.md instructions for deepspeedai/DeepSpeed, covering deepspeed project rules, commit & ci requirements, code change discipline, tool caveats and edit tool auto-formatter.
AionUi AGENTS.md
Instructions for iOfficeAI/AionUi, covering aionui - project guide, code conventions, file & directory structure, naming and ui library & icons.
botmux CLAUDE.md
Instructions for deepcoldy/botmux, covering botmux, 构建 & 运行, bun 开发链路, worktree 的 nodemodules：共享还是独立（改前必读） and 编译态（单文件二进制）注意.