Deep Research Agent Team
A 13-agent pipeline for rigorous academic research on any topic, from vague ideas to a finished APA 7.0 report.
The skill includes user confirmation points (after Phase 1, ethics check), data-flow transparency (treats retrieved content as data, not instructions), sensitive-data handling via ethics review, external effects are mostly retrieval and API calls with graceful degradation (e.g., Semantic Scholar skip). However, permissions are not clearly least-privilege, hook enforcement is referenced but not evidenced, and external API calls lack explicit confirmation mechanisms. Partial deduction.
Instructions are internally consistent, workflows detailed, and failure paths described. But there are no test suites, no CI configuration evidence, reliance on external services (Semantic Scholar, databases, libraries) without demonstrated resilience, and static review cannot verify reproducibility. Hence 6.
Clear audience, scenarios, mode selection flow, multilingual trigger keywords (Chinese, English, Korean), and explicit non-trigger boundaries. However, environment fit for mainland China is unproven; the skill relies heavily on overseas academic databases and Google Scholar which may be inaccessible in China. Partial deduction.
Has version, last updated, status, related skills, and references many files. But lacks install/dependency notes, parameter documentation, examples (beyond minimal), changelog, known limitations, and maintenance responsibility is unclear (publisher unverified). Score 10.
Core task of deep research has detailed pipeline and expected APA 7.0 output format, but no actual output examples or validation evidence; static review cannot confirm direct usability. Score 5.
Key claims (13 agents, 8 modes, PRISMA compliance) lack third-party execution evidence; only documentation exists. No tests, CI, or reproducible experiments provided. Score 3.
- The skill heavily relies on overseas academic databases and tools (e.g., Google Scholar, PubMed, Semantic Scholar) that may be inaccessible or restricted in mainland China, affecting core functionality.
- No permission configuration or least-privilege details provided; runtime may request more permissions than necessary (e.g., hooks, API calls).
- Reference integrity check mentions Retraction Watch but lacks concrete integration or verification steps.
- Publisher is unverified; provenance and long-term maintenance responsibility are unclear, warranting caution.
What does this skill do, and when should you use it?
This is a domain-agnostic deep research skill that orchestrates a team of 13 agents across six phases to guide users from a vague research idea to a complete APA 7.0 research report. The skill offers eight operational modes, including full research, quick brief, literature review, systematic review with PRISMA, fact-checking, three-way literature scan, Socratic guided research dialogue, and optional meta-analysis. Key features include research question formulation, methodology design, systematic literature search, source verification, cross-source synthesis, risk of bias assessment, editorial review, devil's advocate challenges, ethics review, and post-research literature monitoring. The skill is designed for Claude Code and requires user confirmation and human involvement.
The skill runs a six-phase pipeline. In Phase 1 (Scoping), it transforms vague topics into FINER-scored research questions and a methodology blueprint. Phase 2 (Investigation) conducts a systematic literature search and verifies sources using an evidence hierarchy, predatory journal detection, and conflict-of-interest flagging. Phase 3 (Analysis) synthesizes evidence across sources, identifies contradictions, and performs gap analysis. Phase 4 (Composition) drafts a full APA 7.0 report. Phase 5 (Review) runs parallel editorial, ethics, and devil's advocate checkpoints. Phase 6 (Revision) addresses feedback with up to two revision loops. The skill includes eight modes and supports trigger keywords in English, Traditional Chinese, and Korean.
- Use Socratic mode when a researcher needs guidance to clarify a vague research direction.
- Use literature review mode when a scholar has a clear research question and needs a citable review.
- Use review mode when a researcher needs to evaluate a single paper before deciding to cite it.
- Use quick mode when a user needs a fast brief on a topic.
- Use fact-check mode when a user needs to verify the factual accuracy of specific claims.
- Use systematic-review mode when a researcher needs a PRISMA-compliant systematic review with meta-analysis.
What are this skill's strengths and limitations?
- Covers the full research lifecycle from question formation to final report
- Multiple modes adapt to different needs, including guided and quick options
- Built-in quality checks such as evidence hierarchy and devil's advocate challenges
- Multilingual trigger support including English, Traditional Chinese, and Korean
- Optional systematic review and meta-analysis capabilities
- Dependent on Claude Code and requires manual confirmation for some features
- No test suite provided, so reliability is not formally verified
- Requires network access for literature search and source verification
- Complexity may present a learning curve
How do you install this skill?
This skill is part of a monorepo and can be installed via Claude Code plugin: run /plugin marketplace add Imbad0202/academic-research-skills then /plugin install academic-research-skills in Claude Code. Alternatively, copy the deep-research folder into your project's .claude/skills/ directory.
How do you use this skill?
After installation, make a research request in natural language, e.g., 'Research the impact of AI on higher education.' For Socratic mode, say 'Guide my research on the impact of declining birth rates on private universities.' The skill will trigger the appropriate mode based on the input.
How does this skill compare with similar options?
This skill is part of a suite of academic research skills and can integrate with others like academic-paper, but this skill covers only the research phase, not paper writing.