Discover Literature With Guided Research14 skills
13-agent research team with 8 modes covering exploratory discovery, systematic reviews, and fact-checking.
Research Blueprint
Safety checkedA suite of 4 Claude Code skills with 27 modes covering the full research-to-publication pipeline: deep research, paper writing, peer review, and orchestrated integrity verification.
Run this command to deploy the blueprint to your environment.
Runs your complete academic research workflow as one connected 10-stage pipeline instead of scattered disconnected tasks. It integrates research discovery, manuscript writing, peer review simulation, and integrity verification so each stage builds on the prior one with context preserved throughout, removing the friction that blocks researchers from moving work forward at scale.
Once installed, you invoke the pipeline by describing your research goal to the orchestrator, which routes you through 10 stages: (1) Research Question + Methodology (guided exploration or structured brief), (2) Manuscript drafting (outline approval before text), (3) Peer review with multi-perspective evaluation, (4) Revision coaching + re-revision with integrity gates at boundaries, (5) Output finalization (MD, DOCX, LaTeX, PDF with venue-specific AI disclosures), and (6) Process summary with collaboration quality evaluation. Every stage ends in a user confirmation checkpoint; integrity gates at Stages 2.5 and 4.5 are non-skippable. Output: publication-ready manuscript with a Material Passport documenting all decisions and audit trails.
13-agent research team with 8 modes covering exploratory discovery, systematic reviews, and fact-checking.
12-agent paper authoring pipeline with 11 modes for drafting, revising, formatting, and disclosure.
7-agent review panel with 6 modes for peer review simulation, methodology focus, and reviewer calibration.
10-stage orchestrator with Material Passport handoff contracts, integrity gates, and compliance audits.
Every research decision, from RQ framing through revision, stays grounded in prior work instead of starting cold at each phase, so your final paper reflects a chain of evidence rather than disconnected pieces.
Requires Anthropic API key (ANTHROPIC_API_KEY). Cross-model verification uses OpenAI endpoints when ARS_CROSS_MODEL is set. Optional environment flags: ARS_PASSPORT_RESET=1 (context reset at checkpoints), ARS_CLAIM_AUDIT=1 (opt-in claim-faithfulness audit at Stage 4→5, default OFF), ARS_SOCRATIC_READING_PROBE=1 (reading comprehension honesty probe in Stage 1). Citation verification uses Semantic Scholar + OpenAI-compatible endpoints (supports MiMo, DeepSeek, self-hosted) for cross-model checks. Verification cache persists to ~/.cache/ars/verification.db (90-day TTL).
Requires Claude Code v3.7.0+ for plugin installation; traditional symlink installs also supported. Python optional (needed only for revision-patch mode and submission-package verifier). LaTeX PDF compilation needs tectonic plus optional pandoc for DOCX. This is a copilot tool (user as decision-maker), not full automation. Integrity gates explicitly require user judgment at each stage. Long-running pipelines benefit from passport reset boundaries (opt-in ARS_PASSPORT_RESET=1) to work across sessions.
Deep Research (8 modes)
Literature discovery with 13-agent guided research team. Modes include full research reports, systematic reviews (PRISMA), Socratic exploration, fact-checking, and three-way paper comparisons. Optional pre-screened literature corpus input.
Academic Paper Writing (11 modes)
12-agent paper authoring pipeline with Style Calibration, Writing Quality Check, LaTeX hardening, revision coaching, and figure verification. Outputs Markdown, DOCX, and publication-ready PDF (APA 7.0, Chicago, IEEE, MLA).
Peer Review Simulation (6 modes)
7-agent multi-perspective review with EIC + 3 domain reviewers + Devil's Advocate. Produces 0-100 quality rubric scoring, editorial decisions, revision roadmaps, and re-review verification with traceability matrices.
Pipeline Orchestration + Integrity Gates
10-stage choreography with Material Passport handoff contracts, claim verification (100% at Stage 4.5), temporal consistency checks, and compliance audits (PRISMA-trAIce, RAISE). Mandatory integrity gates at Stages 2.5 and 4.5; no skips permitted.
Anti-Hallucination + Trust Layers
Citation locator infrastructure with three-layer anchors (quote/page/section). Optional claim-faithfulness audit matching cited text to assertions. Cross-model verification option (OpenAI + Anthropic). Deterministic reference existence check across 4 bibliographic indexes.