| name | research-add-fields |
| description | Add more field definitions to an existing fields.yaml. Source new fields from user input or from a web-search subagent suggesting common fields in the domain. |
| disable-model-invocation | true |
| allowed-tools | Bash, Read, Write, Glob, WebSearch, WebFetch, Task, AskUserQuestion, mcp__exa_websearch__web_search_exa, mcp__exa_websearch__web_fetch_exa |
Research Add Fields — Extend Field Definitions
Trigger
/ultra-research:research-add-fields
Manual-only via disable-model-invocation: true. Run between /ultra-research:research and /ultra-research:research-deep to enrich the schema the deep-research subagents will populate.
Pipeline Context
Phase 2b of 5 (optional). Edits fields.yaml in place. When finished, Step 4 uses AskUserQuestion to route the user into the next phase (re-run this skill, /ultra-research:research-add-items, or /ultra-research:research-deep) with (Recommended) annotations based on coverage of the two YAML files.
Workflow
Step 1 — Locate Fields File
Glob */fields.yaml in the current working directory and read it. If multiple match, ask which.
Step 2 — Pick Supplement Source
Use AskUserQuestion with two options:
- A. User input — user lists field names + descriptions directly
- B. Web search — see dispatch rule below
⛔ Dispatch-only. If the user picks B (web search), the orchestrator MUST NOT call WebSearch, WebFetch, or mcp__exa_websearch__* directly. Instead, launch one web-search-agent subagent (background, via the harness's Agent/Task tool) with a prompt asking it to propose common fields used in this domain (e.g. for "AI coding tools", fields like context_window, benchmark_scores, pricing_tiers). The orchestrator's job is dispatch + read the subagent's structured return, nothing else.
Step 3 — Confirm, Categorize, Save
- Show the proposed new fields to the user
- For each confirmed field, ask the user which
category it belongs to (offer existing categories from fields.yaml plus "create new") and which detail_level (brief, moderate, detailed)
- Append to
fields.yaml under the chosen category, preserving existing entries
Step 4 — Inspect Coverage and Route to Next Phase
After saving, Read the updated fields.yaml and the sibling outline.yaml. Apply the thinness rubric:
| Check | Trigger for (Recommended) |
|---|
fields.yaml total fields | Still < 6 fields, OR only one field_categories entry, OR no field with detail_level: detailed → recommend Add more fields (run this skill again) |
outline.yaml items count | < 8 items, OR no category field used → recommend Add more items |
If neither file is thin, recommend Start deep research instead.
Use AskUserQuestion with header: "Next step" and these options (append (Recommended) to the label when the rubric flags it):
- Add more items — run
/ultra-research:research-add-items
- Add more fields — run
/ultra-research:research-add-fields again
- Start deep research — run
/ultra-research:research-deep
- Done for now — stop here
In the question body, briefly state why each recommendation was made. Invoke the picked skill immediately via the Skill tool.
Output
{topic}/fields.yaml updated in place, with the appended fields echoed back to the user.
