Skill Claude CodeCodex
CARS (Collaborative Academic Research System) - Modular multi-agent academic research system. Uses Socratic dialogue, literature search, and the chessboard reading method to identify research gaps. Supports multi-topic management, cognitive diagnosis, literature scouting, and gap analysis. Trigger keywords: research…
Skill Claude CodeCodex
Research blueprint compiler module. Converts Supervised Layer 5 convergence insights into structured execution specifications. Bridges ideation phase (dialogue → profile → literature → matrix) and research phase (experiment → code → paper). Trigger: Supervised session completes with convergence insight.
Agent
Research blueprint compiler. Converts Supervised Layer 5 convergence insights into structured execution specifications for downstream research modules (Experiment Agent, Code Agent, Paper Writing Module).
Skill Claude CodeCodex
Literature deep-reading and deconstruction module. Invokes Gemini multimodal capabilities to deep-read PDF papers, outputs structured Markdown reports, then parses them into JSON for downstream Matrix Engine consumption. User-triggered.
Agent
Calls the deepreadpdfs MCP tool in cruncher.py to batch deeply read PDF papers into structured Markdown reports.
Agent
Reads deep-reading Markdown reports output by cruncher.py, parses structured content (including problematization fields), converts it to JSON format (v2.0.0), and stores it in hypotheses/translated/ for downstream Matrix Engine consumption.
Skill Claude CodeCodex
Literature Scout Module. Orchestrates 5 agents to complete the entire literature retrieval process: keyword search / seed recommendation → relevance filtering → batch download. Triggered proactively by the user. Trigger keywords: search papers, find literature, 搜索文献, 找论文, seed recommendation, expand from seeds, 种子推荐…
Agent
Receives the filtered list of paper IDs, calls the Semantic Scholar batch API to download full metadata, saves individual JSON files, downloads open-access PDFs, and maintains a consolidated metadata file.
Agent
Master control agent of the Literature Scout module. Interacts with the user and coordinates the four sub-agents: querybuilder, seedexpander, relevancejudge, and batchfetch to complete the entire literature search process.
Agent
Extracts keywords from userprofile, generates multi-perspective natural language queries, calls the Semantic Scholar API to search for papers, and outputs a deduplicated candidate paper set.
Agent
Reads candidate paper collections, performs confidence scoring based on the match between abstract content and the user's research direction; high confidence automatically triggers download, while medium/low confidence is shown to the user for decision.
Agent
Reads searchseeds from userprofile, calls Semantic Scholar recommendation API to expand relevant literature based on seed papers, deduplicates, and outputs a candidate paper set.
Skill Claude CodeCodex
Matrix Engine module. Reads structured paper JSONs from hypotheses/translated/, extracts problem and method axes, builds a 2D chessboard matrix, identifies research gaps, and outputs matrix.json + matrix.csv. User-triggered.
Agent
Reads structured paper JSONs, extracts problem × method axes, builds a taxonomy-based matrix, identifies research gaps with problematization analysis, outputs matrix.json, and automatically calls MCP tool to export CSV visualization.
Skill Claude CodeCodex
Profile Module. Consolidates Socratic session transcripts into userprofile.json and synchronizes Project README metadata.
Agent
Reads the latest Socratic session and existing userprofile.json, extracts/merges cognitive materials, incrementally updates the global profile, and synchronizes the project README.
Skill Claude CodeCodex
Socratic research guidance module. Four agents collaborate to extract raw cognitive material through multi-round dialogue. Supports cold start (zero-shot) and iterative guidance (with user profile). Outputs dialogue records for Profile Module, Literature Scout, and Matrix Engine. Trigger keywords: guide my research…
Agent
Challenges assumptions at key Socratic dialogue nodes, tests logic chains, and identifies bias. It is a sub-agent called by socraticmentoragent.
Agent
Use Socratic questioning to extract raw cognitive material from the user. Supports both zeroshot (cold start) and supervised (hot start) modes.
Skill Claude CodeCodex
Socratic research guidance module. Four agents collaborate to extract raw cognitive material through multi-round dialogue. Supports cold start (zeroshot) and iterative guidance (supervised). Outputs dialogue records for Profile Module, Literature Scout, and Matrix Engine. Trigger keywords: guide my research, help me…
Agent
After each user reply, decide whether the dialogue should stay, progress, redirect, or regress. Supports both zeroshot and supervised modes with different layer definitions and exit signals.
Agent
Review question quality before the Socratic mentor asks; ensure questions match layer depth and protocol rules for both zeroshot and supervised modes.