| name | research-add-items |
| description | Add more items (research objects) to an existing outline.yaml. Optionally launches a web-search subagent to suggest items in the topic's 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 Items — Extend Item List
Trigger
/ultra-research:research-add-items
Manual-only via disable-model-invocation: true. Run between /ultra-research:research and /ultra-research:research-deep whenever the initial item list missed something.
Pipeline Context
Phase 2a of 5 (optional). Edits outline.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-fields, or /ultra-research:research-deep) with (Recommended) annotations based on coverage of the two YAML files.
Workflow
Step 1 — Locate Outline
Glob */outline.yaml in the current working directory and read it. If multiple match, ask which.
Step 2 — Gather Candidate Items (Two Sources in Parallel)
Use AskUserQuestion to ask both:
- "Which items do you want to add? (free-form list)"
- "Run a web search to suggest more items in this domain? (yes/no)"
⛔ Dispatch-only. If web search is requested, 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 for additional items in the topic's domain, returning each as name: short description. The orchestrator's job is dispatch + read the subagent's structured return, nothing else.
Step 3 — Merge, Deduplicate, Confirm, Save
- Combine user-named items + web-search suggestions
- Deduplicate by
name (case-insensitive)
- Show the proposed additions to the user for confirmation
- Append the confirmed items to
outline.yaml (preserve existing entries) and save
Step 4 — Inspect Coverage and Route to Next Phase
After saving, Read the updated outline.yaml and the sibling fields.yaml. Apply the thinness rubric:
| Check | Trigger for (Recommended) |
|---|
outline.yaml items count | Still < 8 items, OR no category field used → recommend Add more items (run this skill again) |
fields.yaml total fields | < 6 fields, OR only one field_categories entry, OR no field with detail_level: detailed → recommend Add more fields |
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 again
- Add more fields — run
/ultra-research:research-add-fields
- 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}/outline.yaml updated in place, with the appended items echoed back to the user.
