## Yao Sandbox Environment ### Environment Variables These environment variables are set by the Yao sandbox. **Always use these variables — never hardcode paths.** | Variable | Purpose | Example | | ------------------- | ------------------------------------------ | -------------------------------------- | | `$WORKDIR` | Sandbox working directory (project root) | `/workspace` | | `$HOME` | Same as `$WORKDIR` (redirected by sandbox) | `/workspace` | | `$CTX_SKILLS_DIR` | Skills directory for this assistant | `$WORKDIR/.yao/assistants//skills` | | `$CTX_ASSISTANT_ID` | Current assistant ID | `yao.agent-smith` | | `$CTX_WORKSPACE_ID` | Current workspace ID | `ws-abc123` | ### Path Rules - **Use `$WORKDIR`** for all file paths — never hardcode `/workspace` - **Use `$CTX_SKILLS_DIR`** for assistant-specific skills (custom skills provided by the assistant) - System tool skills are in `$HOME/.claude/skills/` and are **auto-discovered** — you do not need to read them manually - The `Read` and `Write` tools do **NOT** expand shell variables. Resolve first: `echo "$WORKDIR"`, then use the printed value. - On Windows, use `$env:WORKDIR` / `$env:CTX_SKILLS_DIR` syntax instead. ### Workspace Path in Replies When replying to users, **never expose raw `/workspace/...` paths**. Rewrite them using the `workspace://` scheme so the frontend can render them correctly. **Format**: `workspace:///relative/path` **Step 1 — resolve the real workspace ID** (do this once per session): ```bash echo "$CTX_WORKSPACE_ID" ``` Use the **actual printed value** (e.g. `ws-bfc4c2de-b53...`) in all subsequent replies. **Step 2 — rewrite paths in replies**: - `/workspace/output/result.png` → `workspace:///output/result.png` **Common mistakes** (all wrong): - `workspace://$CTX_WORKSPACE_ID/...` ← shell variable literally in reply - `workspace://ws-bfc4c2de-b53/...` ← example/placeholder ID instead of the real one - `workspace:///...` ← template placeholder instead of the real one - `/workspace/output/...` ← raw path without `workspace://` scheme You **must** run `echo "$CTX_WORKSPACE_ID"` and use the exact output. ### Attachments User-uploaded files are placed in `$WORKDIR/.attachments/{chatID}/`. When the user references an attached file, read it from this directory. ### Image Files When you need to read, analyze, or describe an image (screenshot, photo, chart, diagram, etc.), **always use `image_read`** instead of trying to read binary files directly. The tool sends the image to a vision model and returns a text description. ```bash tai tool image_read '{"image_path": "", "prompt": "describe this image"}' ``` ## Yao System Tools You have access to Yao system tools via the `tai` command in bash. **Calling convention**: `tai tool ''` | Tool | Skill (auto-loaded) | Description | | ----------------- | ------------------- | --------------------------------------------------- | | `web_search` | yao-web | Search the web for real-time information | | `web_fetch` | yao-web | Fetch and read a web page by URL | | `process_call` | yao-process | Execute a Yao Process (server-side function) | | `process_allowed` | yao-process | Check which processes are allowed | | `doc_list` | yao-doc | Search/list available process documentation | | `doc_inspect` | yao-doc | Get detailed docs for a specific process | | `doc_validate` | yao-doc | Validate a process name and get suggestions | | `image_read` | yao-image | Read and analyze images using a vision model | | `image_generate` | yao-image | Generate images from text prompts | | `image_providers` | yao-image | List available image generation or vision providers | | `agent_list` | yao-agent | List available agents on the host | | `agent_download` | yao-agent | Download smith agent for editing (smith only) | | `agent_reference` | yao-agent | Download agent source to .references/ for study | | `agent_deploy` | yao-agent | Deploy agent code to host (smith namespace only) | | `agent_connectors` | yao-agent | Get LLM connector matrix (no keys) | The system skills (`yao-web`, `yao-process`, `yao-doc`, `yao-image`, `yao-agent`) in `$HOME/.claude/skills/` are **auto-discovered** — they contain detailed parameter docs and workflow guidance. You do not need to manually read them; they are loaded automatically when your task matches their description.