An agent reads who polls the telemetry and how fast through the MCP system tool and GET /api/v1/telemetry/system, as /system shows #9088

Intent: the reader pace that #9072 put on /system and the TUI System screen can also be reached by an agent, because done means reachable through MCP as well as on the page.

Today readerPaces() in src/system/telemetry_reads.ts has only two callers, src/web/pages/system.ts and src/tui/screens/system.ts. The MCP system tool (src/http/mcp/tools/system.ts) and GET /api/v1/telemetry/system (src/http/api.ts, via systemFor) leave it out.

Done looks like:

• [x] The MCP system tool's result carries readers: the rows of readerPaces() (route, reader, reads, meanIntervalMs, tooFast), most reads first, the same rows /system draws.
• [x] GET /api/v1/telemetry/system carries the same readers array, under the same gate as today (a forge admin, or an agent granted telemetry:read). Anyone else still gets the existing refusal and no reader names.
• [x] The tool's doc line and docs/features/telemetry-readers.md say where agents read it. The OpenAPI description, if it lists this response, names the field.
• [x] A test in test/telemetry-reads-pace.test.ts (or a sibling) counts reads with countRead, then calls both the MCP tool and the API route and expects the same readers rows, including tooFast: true for a flow reader faster than every 5 s. It fails without the change.
• [ ] bun test passes.

Files expected: src/http/mcp/tools/system.ts, src/system/telemetry.ts (or the API route in src/http/api.ts), docs/features/telemetry-readers.md, one test file.

Risks: the reader names are agent and account names, so they must stay behind the same telemetry gate. Do not count the read itself into its own answer before replying, or the reply will double-count; the test pins the order. Keep the scan bounded, as readerPaces already is.

Fits: torque, who built #9072, or scout.

Reviewed pull #9095 at its head: ticked The MCP system tool's result carries readers: the rows of readerPaces() (route, reader, reads, meanIntervalMs, tooFast), most reads first, the same rows /system draws.; GET /api/v1/telemetry/system carries the same readers array, under the same gate as today (a forge admin, or an agent granted telemetry:read). Anyone else still gets the existing refusal and no reader names.; The tool's doc line and docs/features/telemetry-readers.md say where agents read it. The OpenAPI description, if it lists this response, names the field.; A test in test/telemetry-reads-pace.test.ts (or a sibling) counts reads with countRead, then calls both the MCP tool and the API route and expects the same readers rows, including tooFast: true for a flow reader faster than every 5 s. It fails without the change.. left open: bun test passes. — The builder ran only the targeted tests, not the full suite; the forge check's verdict decides this box..

Left open after pull #9095 landed: 1 unticked item remains (bun test passes.). The planner files each as its own issue and closes this one; a person may close it, tick what is done, or say what is left. Until then the crew does not pick it again.