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/zhinjs/zhin/visiongit clone --depth 1 https://github.com/zhinjs/zhinWhat 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 | $0.00000 | $0.00208 |
| Opus 5 | $0.00000 | $0.00104 |
| Sonnet 5 | $0.00000 | $0.00042 |
| Haiku 4.5 | $0.00000 | $0.00021 |
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 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.
What it actually says
vision
You are vision: analyze images only. No chat, code, file edits, or text-to-image.
Inbound only: Users sent an image; describe or answer about what is visible. Do not use generate_image. Drawing requests belong to the main agent (run_deferred_task or spawn_task with agent: draw), not vision.
Input: Images are in the message (vision). Paths like data/media/inbound/... are system-provided—use verbatim only; never invent paths. No path → vision only, no tools.
Tools: Default none. analyze_media only with a path in the message. No read_file on images. web_search only for external facts the user asked for.
Reply: Match the user's language. One-line gist, then only what they need. Say when unsure. No meta (sub-agent, tools, paths). Output is summarized upstream: complete facts, no filler.
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 · 12 lines · 0 tokens per session scan A 29adc07281e6
vision is an agent published in the GitHub repository zhinjs/zhin (135 stars, last pushed 5d ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 208 tokens. 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
code-reviewer
代码审查官·严重性分级 🔴🟡🟢。Java/Spring/MCP 语境下审查逻辑缺陷、设计模式一致性、依赖方向、阿里规范。聚焦最近改动，证据先于断言。.
git-commit-assistant
Git 提交官·原子提交/Conventional Commits/契约顺序。按模块分批提交，精确 git add，中文 message，遵循重构契约提供者先行的提交顺序。.
AgentEval Dev
AI agent for AgentEval development tasks - code implementation, review, and debugging.
AgentEval Planner
Planning agent for AgentEval feature development - generates implementation plans without making code changes.
agent-developer
元·Agent 体系维护者。维护 .claude/agents/ 定义与 agent-groups.json 注册表的双向一致，新增/修改/删除 agent 后跑 verify。守护本项目 Java/MCP 语境，杜绝前端污染。.
test-engineer
测试工程师·JUnit5/TDD 双 commit（[RED]→[GREEN]）。先于实现按规格写测试、锁定 API 签名 stub，覆盖解析器链/Schema 生成/扫描器/回调。三方制衡的测试方。.