Research Framework

Prompt-first procedure: Describe the outcome you want in your agent conversation. The agent should select and load the appropriate AIWG assets, explain material changes, request any needed approval, and report verification evidence. Exact commands and flags appear only in the CLI reference.

First time using AIWG? Begin with Install, Connect, and Verify. This guide assumes AIWG is already installed, `all` is deployed for your provider, and the agent-owned context regeneration procedure has connected the agent to this project.

You're doing academic or technical research. Finding papers manually is slow. Tracking what you've read and what it says is a spreadsheet nightmare. When you cite something, you're never sure you're representing the source correctly.

The research framework automates the mechanical parts — discovery, acquisition, summarization, citation management, quality assessment — while keeping you in control of interpretation and synthesis. It prevents hallucinated citations by grounding everything against actual source content.

Deploy it

Use AIWG to complete this documented outcome: Deploy it
Have it inspect the current state, explain the plan, ask before material
changes, and report the result with verification evidence.

The research workflow

discover → acquire → document → cite → assess quality → synthesize → find gaps → archive

Each stage has dedicated commands. Run them in sequence or jump to wherever you are.

Finding papers

/research-discover "transformer attention mechanisms" the limit option 20

Searches academic databases using semantic search, ranks results by relevance and quality, and returns a prioritized list. The `discovery-agent` detects gaps in your existing corpus and surfaces papers you likely haven't seen.

Filter by recency, citation count, or open access:

/research-discover "RLHF alignment techniques" the year-min option 2023 the open-access-only option

Acquiring papers

/research-acquire the from-results option last the output option papers/

Downloads PDFs with Unpaywall integration for open-access versions. Extracts metadata, validates against FAIR principles, and assigns persistent `REF-XXX` identifiers to every paper. You always know how to refer back to a source.

Documenting papers

/research-document papers/attention-is-all-you-need.pdf

Uses RAG-based summarization against the actual PDF content — not from training data. Produces a structured literature note with:

Key findings and contributions
Methodology summary
Limitations acknowledged by the authors
Connections to other papers in your corpus
GRADE quality assessment

This is the core anti-hallucination mechanism. Every claim in the note is grounded against the source text.

Citation management

/research-cite REF-042 the style option apa

Generates properly formatted citations in 9,000+ styles from stored metadata. No manual formatting.

Check all citations in a document for accuracy:

/verify-citations path/to/paper.md

Flags any citation that doesn't match the actual source content in your corpus. Catches misquotes, misattributions, and fabricated page numbers.

Quality assessment

Every paper gets a GRADE score (Grading of Recommendations, Assessment, Development and Evaluations):

High: Systematic reviews, RCTs, large cohort studies
Moderate: Smaller controlled studies, consistent observational evidence
Low: Case reports, expert opinion, early-stage findings

/quality-assess REF-042

When you cite low-quality sources, the framework flags it and suggests appropriate hedging language. You can still use the source — you just do it honestly.

Corpus health

/research-status

Shows corpus statistics: total papers, quality distribution, coverage by topic, papers with incomplete metadata, and citations that need verification. The `corpus-health` skill gives you a snapshot of where your research stands.

Finding gaps

/research-gap "transformer fine-tuning approaches"

Analyzes your corpus against the query topic to identify what's missing — research questions that aren't answered by what you have, contradictions between sources that need resolution, and time periods with thin coverage.

Synthesizing knowledge

/research-document the synthesize option "attention mechanisms"

Produces a synthesis note connecting related papers in your corpus — agreements, contradictions, evolution of the idea over time. This is where individual literature notes become connected knowledge.

Archiving the corpus

/research-archive the verify-integrity option

Packages the research corpus following OAIS standards: SHA-256 integrity verification, provenance tracking for all derivations, version control for notes, and reproducibility validation so anyone can reconstruct your research process.

Key references

`/corpus-health` — Corpus health report
`/research-status` — Full research status dashboard
`/grade-report` — GRADE quality distribution across corpus
`@agentic/code/frameworks/research-complete/README.md` — Full framework documentation