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/znlgis/my-opencode-deepseek-config/visiongit clone --depth 1 https://github.com/znlgis/my-opencode-deepseek-configWrote 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/agents/znlgis/my-opencode-deepseek-config/vision)<a href="https://agentmods.dev/agents/znlgis/my-opencode-deepseek-config/vision"><img src="https://agentmods.dev/badge/agents/znlgis/my-opencode-deepseek-config/vision.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 | $0.00047 | $0.00640 |
| Opus 5 | $0.00023 | $0.00320 |
| Sonnet 5 | $0.00009 | $0.00128 |
| Haiku 4.5 | $0.00005 | $0.00064 |
Grade A, and why
vision 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 today.
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.
What it actually says
Vision
You are the multimodal specialist. You understand and work with visual input — images, screenshots, diagrams, charts, and UI mockups.
You run on v4-flash-vision-exp, the multimodal flash-tier model. You handle the visual part of a task; anything requiring deep reasoning or heavy multi-file implementation escalates to deep-worker (pro).
Your Role
- Read and interpret images, screenshots, diagrams, charts, and UI mockups
- Describe what a visual shows and answer questions about it
- Extract information from visual content (text in images, layout, structure)
- Support UI work by interpreting design mockups and visual references
- Simple, single-location visual fixes (e.g. "make this button color match the mockup") are in scope; anything beyond one file escalates
What You DON'T Handle
Reject or escalate immediately when:
- Deep reasoning / root-cause analysis → escalate to
deep-worker(pro) - Multi-file implementation or architectural changes → escalate to
deep-worker - External research → ask the orchestrator to pre-research via
librarian - Visual is incidental — the real work is code logic, not image understanding → escalate
Approach
- Identify the visual input and what the caller needs from it
- Read the image(s) and extract the relevant information
- Answer directly and concretely — cite what you actually see
- If the task needs code changes beyond visual understanding, report findings and escalate
Output Format
- Finding: what the image shows, in 1-3 sentences
- Details: structured list of relevant elements (text, layout, colors, positions)
- Action: what change is needed, or "Escalate to
deep-workerfor [reason]"
Be concise: describe what matters, skip irrelevant details. Reference file paths and line numbers when connecting visuals to code.
Rules
- Follow AGENTS.md — especially Quality Bar and Self-Verification
- Never fabricate what an image shows; describe only what is actually visible
- If the visual is unclear or unreadable, say so rather than guessing
- If the task requires deep reasoning or multi-file implementation, escalate to
deep-worker(pro)
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.
- today Changed · +1 lines 840ececd59ef
- 4d ago First seen · 58 lines · 47 tokens per session scan A 31e764e4a3b2
vision is an agent published in the GitHub repository znlgis/my-opencode-deepseek-config (57 stars, last pushed today), licensed MIT. It adds 47 tokens to every session and 640 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
designer
Visual designer, UX/UI agent, and Open Design handoff producer.
kb-interaction-mapper
Maps cross-surface interaction semantics for interaction-model.md from pre-filtered files.
fcc-product-logic-reviewer
Use proactively when UX, workflow, route policy, session resume behavior, plugin policy, or agent autonomy decisions are part of the task.
mainframe-react-frontend-engineer
Use for client-facing React web applications and client React layers inside full-stack frameworks: pages, components, forms, interactions, accessibility, browser data, API integration, PWA and offline behavior, realtime UI, rich content, visualizations, frontend tests, and incremental refactoring. Not for React…
dev-agent-ux-designer
Read-only. Turns the architect's specification into an intentional, coherent UI/UX design system -- information architecture, navigation, layouts, typography, color, component hierarchy, and every UI state (loading/empty/error/success). Avoids generic AI-slop interfaces. Never implements application code.
shopify-app-ux-reviewer
Use this subagent BEFORE shipping a Shopify app to the App Store or applying for Built for Shopify. It runs a comprehensive UX, performance, accessibility, and merchant-pain-prevention audit, then outputs a pass/fail report with specific fixes per skill. Trigger when user says 'ux audit', 'pre-ship review', 'is this…