* refactor(usage): remove the unconsumed per-agent USAGE_UPDATE event
USAGE_UPDATE was published per active agent each sweep, bridged, and
broadcast to /ws/system, but no panel client ever consumed it — the
dashboard reads only the aggregate USAGE_SNAPSHOT. Every emission was
wasted event-bus and WebSocket traffic.
Drop the UsageUpdate payload, publish_usage_update and its throttle, the
EventType member, and the bridge subscription. Keep USAGE_SNAPSHOT, which
already carries the per-agent breakdown, so no live data is lost.
* refactor(prompter): remove the legacy local-LLM HTTP endpoints
The panel uses only the live SDK-intake path (/prompter/live/*); the legacy
/prompter/chat, /draft and /sessions/* endpoints — backed by the local Ollama
LLM with hardcoded prompts — had no remaining caller. Remove the router, its
mount in app.py, and its integration test. The live router and the shared
draft-confirmation service are untouched.
* refactor(prompter): drop the dead legacy local-LLM service + schemas
With the legacy HTTP endpoints gone, the local-LLM chat/draft/session methods,
their prompt constants, the ConfirmOverrides/TurnResult dataclasses, and the
entire prompter schema module had no production caller (only their own tests).
Remove them, keeping the live-intake path: create_task_from_draft /
confirm_live_draft, the enum/priority/team coercion, and the pure
description/readiness helpers.
* refactor(agents): stop granting the Task sub-agent tool to roles
Every agent role was granted the built-in Task tool, but no role prompt or
workflow uses it and there are no custom sub-agent definitions — so a Task call
only spawns a context-blind generic sub-agent that burns budget (ToolSearch,
the comment's stated use, is MCP-only and not callable in agent containers).
Drop Task from all three grant points in lockstep: the --tools spawn flag and
both _ROLE_BUILTIN_TOOLS maps (system-prompt + briefing layers), with a
regression guard added to each layer's test.
---------
Co-authored-by: Renn F <rennf93@users.noreply.github.com>
The system-prompt directive layer and the briefing block both opened
with "FIRST ACTION REQUIRED: run ToolSearch to activate deferred
Edit/Write". That premise is false: per Claude Code 2.1.114, ToolSearch
gates only deferred MCP tools, never built-ins — and it is not even a
callable tool in the agent runtime. Built-ins are loaded at spawn via
the `--tools` flag and gated solely by the per-role permission rules
(the actual Edit/Write breakage was the global Write(*)/Edit(*) deny +
single-slash path, fixed in c0ba335). So weak models dutifully chased a
nonexistent ToolSearch, concluded Edit/Write were unavailable, and
rewrote whole files via destructive shell redirection.
Both touch points now affirm the role's built-in tools are loaded and
ready, tell the agent NOT to call ToolSearch, and (for authoring roles)
explicitly steer away from whole-file shell redirection — directly
countering the clobber behaviour. Role prompt files (developer,
cell_pm, main_pm, board) updated to match. Dead
_read_tool_load_from_role_prompt (no callers) removed. Directive tests
rewritten to lock the corrected behaviour.
Smoke-8 follow-up. Two issues in _write_agent_briefing:
1. _build_tool_load_block was scraping role prompts for a "## Load on
spawn" section that doesn't exist in any role file. Returned "" for
every role → no ToolSearch directive in the briefing. Combined with
weak models skipping the system-prompt-layer directive (#144), the
agent's first action was Edit → "not enabled in this context."
Fix: per-role tool list lives in the orchestrator (mirrors
factories._base.py). Pre-renders the directive directly. developer
and documenter get Edit + Write; QA/PMs/board get the common
read-only set. 7 tests pin the contract.
2. The briefing's "Terminal tools (how to exit cleanly)" section still
listed pre-gateway verb names: roboco_agent_idle,
roboco_task_substitute, roboco_task_escalate,
roboco_task_submit_qa, _qa_pass/fail, _docs_complete, _complete.
Same rename pattern as #145's _TERMINAL_TOOLS set. Updated to:
i_am_idle, i_am_blocked, unclaim, i_am_done, pass, fail,
i_documented, complete, submit_up, escalate_up, escalate_to_ceo.
The agent now reads the same directive in two places (system prompt +
session briefing) — the second touch point catches weak models that
skip the first.
