moonlight-lupin/agent-skills / research/deep-research
deep-research
VerifiedUse when the user asks for "deep research", a "research report on", a "comprehensive analysis of", to "look into X in depth", to "write a report on X", or any question needing multi-source synthesis beyond a single search — an autonomous multi-step deep research engine implementing an iterative Think → Search → Extract → Synthesize → Stop loop. The LLM drives every decision: what to search, what's relevant, what's missing, and when to stop. Produces a cited, magazine-quality report with inline citations, category-specific formatting, and research stats. For entity vetting/dossiers use entity-research; for news digests use news-monitoring; for source-grounded Q&A use notebooklm-mode.
evals/routing-fixtures.json
View on GitHub{
"skill_name": "deep-research",
"purpose": "Lightweight routing + output-contract fixtures. These are specs, not executed against a live model in CI. tests/test_routing_fixtures.py validates that each fixture is well-formed and routes to a real skill; they also serve as worked examples of when deep-research is (and is NOT) the right skill, and what its output must contain.",
"fixtures": [
{
"id": "dr-01-cited-report",
"request": "Write a thorough, cited report on the state of solid-state battery commercialization in 2026 — the key players, expected timelines, and the remaining technical barriers.",
"expected_routing": {
"skill": "deep-research",
"reason": "Open-ended multi-source synthesis into a single cited report. Not a recurring digest, not a dossier on one named entity, not Q&A over user-supplied sources.",
"not": ["news-monitoring", "entity-research", "notebooklm-mode"]
},
"required_output_fields": ["numbered inline citations [N] matching the source table", "evidence-basis tags ([VERIFIED]/[SOURCED]/[REASONED]/[ESTIMATED]) with an evidence key", "sources/references list with quality tiers", "research stats"],
"forbidden_patterns": ["(?i)as an ai language model", "(?i)i cannot (browse|access) the (web|internet)", "(?i)\\bi made (this|that|these) up\\b"]
},
{
"id": "dr-02-boundary-fact-check",
"request": "Is it true that more than half of new cars sold in Norway in 2025 were fully electric? Check this and show your sources.",
"expected_routing": {
"skill": "deep-research",
"reason": "Since fact-checker retired (its claim-verification role absorbed into deep-research), a single, specific claim to verify routes to deep-research: its fact-check report category, evidence-basis gates, and citation discipline cover claim-level verification.",
"not": ["news-monitoring"]
},
"required_output_fields": ["evidence for/against two-column table", "preliminary verdict", "evidence-basis tags ([VERIFIED]/[SOURCED]/[REASONED]/[ESTIMATED]) with an evidence key", "numbered inline citations [N] matching the source table"],
"forbidden_patterns": ["(?i)as an ai language model", "(?i)i don't have access to", "(?i)\\bverdict: false\\b"]
},
{
"id": "dr-03-boundary-dossier",
"request": "Give me a background dossier on Acme Robotics Ltd — who owns it, any adverse media, and whether it has sanctions exposure.",
"expected_routing": {
"skill": "entity-research",
"reason": "Background/vetting of one named entity with ownership, adverse-media and sanctions signals is entity-research, not a general research report.",
"not": ["deep-research"]
},
"required_output_fields": ["entity identity", "ownership/management", "adverse media", "sanctions/PEP signals", "sources"],
"forbidden_patterns": ["(?i)\\bcleared\\b", "(?i)\\bno sanctions risk\\b", "(?i)compliance determination"]
},
{
"id": "dr-04-boundary-vault",
"request": "I'm going to paste five PDFs. Answer my questions strictly from those documents, with quotes, and tell me when they don't cover something.",
"expected_routing": {
"skill": "notebooklm-mode",
"reason": "Strictly-grounded Q&A over user-supplied sources is notebooklm-mode (vault), not open-web deep research.",
"not": ["deep-research"]
},
"required_output_fields": ["grounded answer", "verbatim quotes", "source attribution", "gaps identified"],
"forbidden_patterns": ["(?i)as an ai language model"]
},
{
"id": "dr-05-clarify-vague",
"request": "can you research AI for me please",
"expected_routing": {
"skill": "deep-research",
"reason": "Routes to deep-research, but the in-skill Step 0 clarification phase should trigger: the request is too vague (2 words, no scope) and should elicit a clarification before planning. The fixture asserts the skill is correct but the first agent output should be a clarification question, not a report.",
"not": ["news-monitoring", "entity-research", "notebooklm-mode"]
},
"required_output_fields": ["clarification question(s) before any research begins"],
"forbidden_patterns": ["(?i)as an ai language model"]
},
{
"id": "dr-06-skip-clarification-specific",
"request": "Write a comprehensive comparison of Qdrant vs pgvector for RAG applications — performance, scalability, ease of deployment, and cost.",
"expected_routing": {
"skill": "deep-research",
"reason": "Specific multi-faceted comparison question with clear scope. Step 0 clarification should NOT trigger — the question is specific enough to generate productive search queries immediately. Routes to deep-research and proceeds straight to planning.",
"not": ["news-monitoring", "entity-research", "notebooklm-mode"]
},
"required_output_fields": ["numbered inline citations [N] matching the source table", "evidence-basis tags ([VERIFIED]/[SOURCED]/[REASONED]/[ESTIMATED]) with an evidence key", "comparison table", "sources/references list with quality tiers", "research stats"],
"forbidden_patterns": ["(?i)as an ai language model", "(?i)before i begin researching, i'd like to clarify"]
}
]
}
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| deprecation-and-migrationaddyosmani/agent-skillsManages deprecation and migration. Use when removing old systems, APIs, or features. Use when migrating users from one implementation to another. Use when migrating a database schema in production, such as renaming or dropping a column without downtime (expand/contract). Use when deciding whether to maintain or sunset existing code. | Stars: 119k | Updated | Claude CodeCodex | Verified | Install |
| performance-optimizationaddyosmani/agent-skillsOptimizes application performance across frontend, backend, queries, and databases. Use when performance requirements exist, when you suspect performance regressions, when Core Web Vitals or load times need improvement, when N+1 query patterns need fixing, or when profiling reveals bottlenecks. | Stars: 119k | Updated | Claude CodeCodex | Verified | Install |
| agent-cost-reportthedotmack/claude-memBelievable agent cost report for any period, default the last 7 full days PT, not counting today. Measured tokens from Claude Code transcripts priced at OpenRouter list prices (ESTIMATED), measured provider spend when a sanctioned source exists, note-taker cost separate, Timing-style HTML/PDF plus report.json, line-items.csv, evidence.json. | Stars: 118k | Updated | Claude CodeCodex | Verified | Install |
| agent-cost-reportthedotmack/claude-memBelievable agent cost report for any period, default the last 7 full days PT, not counting today. Measured tokens from Claude Code transcripts priced at OpenRouter list prices (ESTIMATED), measured provider spend when a sanctioned source exists, note-taker cost separate, Timing-style HTML/PDF plus report.json, line-items.csv, evidence.json. | Stars: 118k | Updated | Claude CodeCodex | Verified | Install |