| name | rumil-dispatch |
| description | Fire one rumil call (find_considerations, assess, scout-*, web_research, create_view) as Claude Code. This is the "rumil-mediated" lane — a normal rumil call with its usual context-building, prompts, and tools. Use when the user wants to investigate a question deeper, assess a judgement, run a specific scout, or produce a distilled view summary. The run is tagged with origin=claude-code in the trace. |
| allowed-tools | Bash |
| argument-hint | <call_type> <question_id> [--budget N] [--smoke-test] |
rumil-dispatch
Fires one rumil call and streams terse progress into the conversation.
This is the rumil-mediated lane: the call is a standard rumil call
with its own context-building and prompts — Claude Code is just the
trigger. Distinguishable from a main.py-initiated run via the
origin=claude-code tag in runs.config and calls.call_params.
Defaults
- Budget:
--budget 3 by default. Match this to the work you're asking
for — single assess/scout calls don't need more. If the user wants a full
multi-call investigation, suggest they use main.py --continue instead,
which the orchestrator can spread budget across.
- Workspace: reads active workspace from
.claude/state/rumil-session.json.
Override with --workspace <name> if needed.
- Visibility: the script prints the trace URL immediately. Surface it to
the user in your response so they can open it in a browser alongside the
CC session.
Call type picking
If the user asked for a specific type, use it. Otherwise:
- "investigate X more" / "find considerations" →
find-considerations
- "assess X" / "judge X" / "how credible" →
assess
- "brainstorm subquestions" →
scout-subquestions
- "look for evidence on the web" →
web-research
- "summarize what we know" / "distill a view on X" →
create-view (produces or updates a curated View summary page for a question; appropriate once a question has enough research to warrant a distilled summary)
When in doubt, ask the user before firing. Dispatching is not free.
Invocation
setopt no_glob 2>/dev/null; set -f; PYTHONPATH=.claude/lib uv run python -m rumil_skills.dispatch_call $ARGUMENTS
After it runs
The script prints:
- trace URL (surface this)
→ firing <call_type> (budget N) — confirmation line✓ done: status=... cost=$... budget=used/total — completion line- the call's
result_summary if non-empty
If the call failed or the result looks off, suggest using /rumil-trace <call_id>
to inspect what happened. The short call ID is in the trace URL.
