deep-research
/deep-research
Bada nieznane technologie i zgodność frameworków oraz bibliotek, weryfikuje sygnatury API i zestawia dowody techniczne.
- Autor
- hmfareed ↗
- Repozytorium
- Frontier-Engineer ↗
- Licencja
- MIT
- Ostatnia zmiana repozytorium
- 13.11.2026
Ocena i bezpieczeństwo
- Pliki w paczce
- 2
- Skrypty
- 0
- Komentarze HTML
- 0
- Adresy URL w SKILL.md
- 0
Brak sygnałów. Repozytorium zmienione 13.11.2026, bez odwołań do wycofanych modeli i API.
MIT, otwarta: można używać i zmieniać, z zachowaniem informacji o autorze.
Co robi i kiedy
Bada nieznane technologie i zgodność frameworków oraz bibliotek, weryfikuje sygnatury API i zestawia dowody techniczne.
Oryginalny opis (EN)
Use when investigating unfamiliar technologies, researching framework/library compatibility, verifying API signatures, evaluating technical alternatives, cross-referencing security advisories, or synthesizing technical evidence.
Jak działa
- Purpose
- Activation criteria
- Non-activation guidelines
- Required context and inputs
- Operating procedure
- Technical and engineering standards
- Verification and evidence requirements
- Expected outputs and deliverables
- Risks, boundaries and failure modes
- Progressive disclosure references
Podgląd SKILL.md
--- name: "deep-research" description: "Use when investigating unfamiliar technologies, researching framework/library compatibility, verifying API signatures, evaluating technical alternatives, cross-referencing security advisories, or synthesizing technical evidence." --- # Deep Research — Technical Investigation and Authoritative Evidence Synthesis > Generated from `specs/skills/deep-research.md` by `scripts/generate-antigravity-skills.ts`. Edit the canonical specification, not this file. ## Purpose The Deep Research skill provides a disciplined methodology for investigating unfamiliar technologies, complex APIs, library compatibility, and technical trade-offs. It ensures that engineering decisions are grounded in verified, authoritative primary sources, current version specifications, and explicit trade-off analyses, preventing reliance on stale forum advice or hallucinated capabilities. ## Activation criteria Activate this skill when: - Investigating an unfamiliar framework, library, API, or database technology. - Resolving dependency compatibility conflicts, peer dependency warnings, or breaking changes. - Evaluating multiple competing technology choices for a new project or subsystem. - Assessing security advisories, CVE reports, or regulatory requirements. - Determining whether an official platform API supports a requested feature. ## Non-activation guidelines Do not activate this skill when: - Implementing already-decided features using existing codebase conventions. - Debugging local syntax errors or simple runtime exceptions (use Advanced Debugging). - Authoring component unit tests for known interfaces. - Formatting code or modifying visual layouts. ## Required context and inputs Before initiating research: 1. **The Research Objective:** Specific technical question or hypothesis to prove or disprove. 2. **Current Project Environment:** Runtime versions, framework versions, operating system, and target deployment platform. 3. **Constraints:** Licensing requirements (MIT, Apache 2, GPL limits), package size limits, security boundaries, and operational budget. 4. **Authoritative Sources:** Official documentation portals, repository URLs, and release note locations. ## Operating procedure Execute the following ten-step technical research workflow: 1. **Define the Research Question:** Formulate a precise, bounded technical question (_"Does Next.js 15 support WebSocket streaming in App Router Route Handlers natively?"_). 2. **Identify What Must Be Verified:** List concrete facts needed (API signatures, runtime support, breaking changes, performance bounds). 3. **Prioritize Official Documentation:** Consult Tier 1 primary sources (MDN, official language/framework docs) first. 4. **Search for Current and Relevant Sources:** Ensure consulted sources match the exact major version used in the project. Reject outdated blog posts from superseded versions. 5. **Compare Competing Approaches:** If evaluating alternatives, analyze at least two viable solutions across performance, maintainability, community health, and license terms. 6. **Verify Version Compatibility:** Inspect `package.json` peer dependencies, node engine ranges, and native binary prerequisites. 7. **Identify Conflicting Information:** If sources disagree, cross-reference against official GitHub issue trackers, commit logs, or run a local reproduction script to verify ground truth. 8. **Distinguish Facts from Assumptions:** Label statements as verified facts, plausible inferences, or unverified assumptions. 9. **Summarize Findings:** Synthesize findings into clear technical prose with executive summary and detailed trade-off analysis. 10. **Provide Recommendations and Citations:** Conclude with an actionable recommendation, citing exact URLs, documentation chapters, and dates checked. ## Technical and engineering standards - **Source Priority Rule:** Strictly respect the source hierarchy: Official Docs > Official Repos & Release Notes > Standards & Security Advisories > Reputable Tech Blogs > Community Forums. - **Never Fabricate:** Never invent API parameters, method names, benchmarks, or compatibility claims. If an API is undocumented or unclear, label it explicitly as unverified. - **Temporal Anchoring:** Always state the version and date of consulted evidence. - **Reproducible Evidence:** Whenever evaluating an unclear edge-case behavior, write a small local proof-of-concept test script to establish definitive empirical evidence. ## Verification and evidence requirements - Cite specific documentation sections, URLs, or official changelog entries for all factual assertions. - Explicitly state which version of the software was inspected. - State explicitly if any question could not be verified due to lack of public documentation or access. ## Expected outputs and deliverables - **Technical Research Report:** Structured briefing containing executive summary, methodology, detailed findings, and recommendations. - **Alternative Comparison Matrix:** Table comparing options across key technical dimensions. - **Compatibility & Migration Advisory:** Clear warning of breaking changes, peer dependency constraints, or deprecation notices. - **Source Citations:** Formatted list of authoritative references. ## Risks, boundaries and failure modes - **Stale Information Hazard:** Applying advice written for an obsolete version that causes silent breakage or security flaws. - **Marketing Bias:** Believing landing page claims of library authors without verifying open GitHub issues or real-world benchmarks. - **Unverified AI Hallucination:** Generating fictional API signatures because they sound plausible. ## Progressive disclosure references For source hierarchy standards, fact vs. assumption criteria, and verification templates: - [Deep Research: Evidence Synthesis, Source Verification, and Fact-Checking Guide](./references/evidence-synthesis-guide.md)
Pliki w paczce (2)
- SKILL.md 6 KB
- references/evidence-synthesis-guide.md 3 KB
Wersje
| Repozytorium | Wynik | Licencja |
|---|---|---|
| Ta strona · hmfareed/Frontier-Engineer | 90 | MIT |
| shepherdjerred/monorepo | 100 | GPL-3.0 |
| 3x-haust/ondra | 100 | MIT |
| glebis/claude-skills | 100 | MIT |
| drvoss/everything-copilot-cli | 100 | MIT |
| 5dive-ai/skills | 100 | MIT |
| huayu-pro/skills | 100 | MIT |
| lovstudio/skills | 100 | MIT |
| bytedance/deer-flow | 100 | MIT |
Migawka bazy z 13.11.2026 · wersja paczki 3c2760546fd3 (skrót SHA-256 całej paczki)
Pobrania w ostatnich 30 dniach: 0 · ranking
Zgłoś problem z tym skillem: szkodliwa treść, błędny opis albo prośba autora o usunięcie.
Podobne skille
- 100technical-research
Weryfikuje twierdzenia techniczne, porównuje opcje i bada nieznane biblioteki, API i narzędzia, z cytowanymi dowodami i oceną pewności. - 100research
Prowadzi pogłębiony research rozwiązań technicznych: przeszukuje sieć, analizuje repozytoria GitHub i zbiera dowody o opcjach wdrożenia lub technologiach. - 100looking-up-docs
Znajduje dokładną dokumentację bibliotek, API i frameworków w zgodnej wersji jednym przepływem wyszukiwania: sygnatury API, klucze konfiguracji, składnia i przykłady, także przez Context7. - 90research
Bada wskazaną bibliotekę, framework, SDK, API lub usługę albo szuka przewodnika migracji, wysyłając agenta zwiadowcę. - 95ark-research
Bada rozwiązania techniczne przez przeszukiwanie sieci, analizę repozytoriów GitHub i zbieranie dowodów, przy ocenie opcji implementacji i technologii. - 100research
Prowadzi pogłębiony research techniczny z oceną wielu opcji przy wyborze technologii, decyzjach architektonicznych, porównaniu bibliotek oraz analizie wydajności.