Ask a natural-language QUESTION about this repository / machine and get a deep, grounded answer instead of just running work. Use this for 'how do I test STT/TTS?', 'where does auth get wired?', 'why does the relay fall back?' — anything where the user wants understanding, not a change. It spawns...
AI agents call yaver_ask to retrieve information from Yaver without modifying anything — typically the context-gathering step in research, monitoring, and reporting workflows, before the agent takes action elsewhere.
| Parameter | Type | Required | Description |
|---|---|---|---|
depth | string | — | How hard to analyze. 'single' = one read-only agent (fast). 'deep' = a multi-agent graph (investigate → answer → verify) for broad/architectural questions. 'aut |
model | string | — | Model id forwarded to the runner. Empty = runner default. |
runner | string | — | Runner ID — claude / codex / opencode. Empty = agent default. |
question | string | Yes | The natural-language question to answer against this repo / machine. |
work_dir | string | — | Optional project directory to scope the analysis to. Empty = the agent's default workdir / auto-detected from the question. |
Parameters from the server's own tool schema.
Even though yaver_ask only reads data, uncontrolled read access leaks sensitive information and racks up API costs — an agent caught in a retry loop can make thousands of calls a minute without anyone noticing.
Attacks that exploit this kind of access
Ask a natural-language QUESTION about this repository / machine and get a deep, grounded answer instead of just running work. Use this for 'how do I test STT/TTS?', 'where does auth get wired?', 'why does the relay fall back?' — anything where the user wants understanding, not a change. It spawns a coding agent that reads the actual code (greps, opens files, follows the wiring), answers with file:line citations, automatically escalates from a shallow scan to a wider cross-checked read when the question is broad or architectural, and explains FIRST — it only modifies the working tree / deploys / touches git after confirming with the user via yaver_ask_user. Returns a task object; stream it with the task's /output SSE or poll get_task for the answer. Prefer this over create_task whenever the intent is to explain rather than to build. It is categorised as a Read tool in the Yaver MCP Server, which means it retrieves data without modifying state.
yaver_ask accepts 5 parameters: depth, model, runner, question, work_dir. Required: question. The full parameter table on this page comes from the server's own tool schema.
Register the Yaver MCP server in PolicyLayer and add a rule for yaver_ask: allow, deny, rate-limit, or require approval. Point your MCP client at the PolicyLayer proxy URL and the rule is enforced on every call, before it reaches Yaver. Nothing to install.
yaver_ask is a Read tool with low risk. Read-only tools are generally safe to allow by default.
Yes. Add a rate_limit block to the yaver_ask rule in your PolicyLayer policy. For example, setting max: 10 and window: 60 limits the tool to 10 calls per minute. Rate limits are tracked per agent session and reset automatically.
Set action: deny in the PolicyLayer policy for yaver_ask. The AI agent will receive a policy violation error and cannot call the tool. You can also include a reason field to explain why the tool is blocked.
yaver_ask is provided by the Yaver MCP server (yaver-cli). PolicyLayer sits as a proxy in front of this server to enforce policies before tool calls reach the server.