Skip to content

Agent Catalog

Raghav Kattel edited this page Jun 1, 2026 · 1 revision

Agent Catalog

Sisyphus Academica has 20+ agents organized into 5 groups. Each is a single Markdown file with frontmatter configuration.

Orchestrator

Agent File Role
Research Director orchestrator/research-director.md Conducts the full 10-phase pipeline. Deploys subagents, collects results, loops on failure. Never writes — only commands.

Subagents (Pipeline)

Agent File Role
Literature Scout subagents/literature-scout.md Multi-source literature search (arXiv, Semantic Scholar, CrossRef, OpenAlex)
Gap Analyzer subagents/gap-analyzer.md Identifies genuine research gaps from the literature corpus
Methodology Designer subagents/methodology-designer.md Recommends statistical tests, power analysis, confound controls
Data Engineer subagents/data-engineer.md Writes analysis code, generates publication-ready figures
Writer subagents/writer.md Writes paper sections with 41 Humanizer patterns as hard constraints
Verifier subagents/verifier.md 3-module verification: citations, statistics, AI-patterns
Style Auditor subagents/style-auditor.md Final Humanizer certification gate
Formatter subagents/formatter.md LaTeX template matching, BibTeX, PDF compilation

Novelty Engines (Parallel)

Agent File Thinking Mode
The Contrarian novelty-engines/contrarian.md Inverts field claims → 10 counter-hypotheses
The Cross-Pollinator novelty-engines/cross-pollinator.md Imports solutions from 15 distant fields
The Assumption Excavator novelty-engines/assumption-excavator.md Finds unstated assumptions, tests if false
The Counterfactual Generator novelty-engines/counterfactual-generator.md Rewrites field history without key papers
The Paradox Sifter novelty-engines/paradox-sifter.md Cross-references "Limitations" sections for contradictions
The Heretic novelty-engines/heretic.md Crown jewel — generates 50 wild hypotheses from title+abstract alone

Adversarial Reviewers (Parallel)

Persona File Focus
Theorist reviewers/theorist.md Formal proofs, mathematical rigor
Empiricist reviewers/empiricist.md Experimental design, baselines
Pragmatist reviewers/pragmatist.md Practical applicability
Skeptic reviewers/skeptic.md "Your results are wrong" — demands proof
Historian reviewers/historian.md Prior art, citation accuracy
Methodologist reviewers/methodologist.md Statistical correctness
Ethicist reviewers/ethicist.md Societal implications
Competitor reviewers/competitor.md Novelty relative to existing work
Student reviewers/student.md Clarity and accessibility
Dreamer reviewers/dreamer.md "What if you went further?"

Skills

Skill File Purpose
Academic Humanizer skills/skill-academic-humanizer.md 11 academic-specific AI-pattern detection patterns (#31-41)
Humanizer (external) ~/.config/opencode/skills/humanizer/ 30 general AI-writing patterns (installed via install.sh)

Clone this wiki locally