Research Agora

Personalized Research Agora setup via scripts/onboard.py

Checkpoint-gated interactive brainstorming for research and knowledge-organization decisions: framing a problem, exploring the option space, and choos...more

Checkpoint-gated interactive brainstorming for research and knowledge-organization decisions: framing a problem, exploring the option space, and choosing a direction WITH the user rather than for them. Use when asked to "brainstorm", "think through options", "explore approaches", "help me decide", "what's the best way to structure/represent/organize X", "weigh alternatives", "design a study/benchmark/ taxonomy", or when a task's hardest part is a decision the user must own. Supports optional multi-agent debate (opposing-prior proposers + a fact-checking devil's advocate) between checkpoints. For software-implementation design and specs, defer to superpowers:brainstorming if that plugin is installed — this skill owns research and representation decisions, not code architecture.less

Find the right skill, or find out what changed. Use when asked "which skill should I use", "what skill do I need", "recommend a skill", "what can the ...more

Find the right skill, or find out what changed. Use when asked "which skill should I use", "what skill do I need", "recommend a skill", "what can the Agora do for", "what should I run first", "quick start", "give me a demo", "get me started", "what's new", "what changed since I last used this", "catch me up", or "changelog". Three modes: **route** maps a task description to a skill, **start** scans the current directory and picks the single highest-value thing to run right now, **changes** reports what was added, deprecated and removed since a date.less

Diagnose abstracts for ML conference papers against structure, venue word limits, specificity, and claim support. Use when asked to "audit my abstract...more

Diagnose abstracts for ML conference papers against structure, venue word limits, specificity, and claim support. Use when asked to "audit my abstract", "diagnose abstract", "check my abstract", "review my abstract", "is my abstract too long", or "abstract feedback". Scores the five-part structure, flags vague or unsupported claims, and returns prioritized fixes. It does not write abstracts.less

Use this agent to detect voice inconsistency across chapters, blog posts, or documents. Activates when asked to "check voice consistency", "tone drift...more

Use this agent to detect voice inconsistency across chapters, blog posts, or documents. Activates when asked to "check voice consistency", "tone drift", "does this sound like me", "voice fingerprint", or "style consistency check". Quantifies rhythm, formality, person, and metaphor density to flag unintentional drift.less

Diagnose root causes of bad writing at the paragraph level. Use when asked to "diagnose this paragraph", "why does this suck", "what's wrong with this...more

Diagnose root causes of bad writing at the paragraph level. Use when asked to "diagnose this paragraph", "why does this suck", "what's wrong with this", "writing diagnosis", or "debug my writing". Identifies patterns like cognitive overload, buried ledes, idea soup, and monotonous rhythm — then teaches the fix.less

Quantitative writing quality verification for scientific papers. Use when asked to "score my writing", "grade this paper", "writing quality check", "v...more

Quantitative writing quality verification for scientific papers. Use when asked to "score my writing", "grade this paper", "writing quality check", "verify writing quality", "how good is my writing", "rate my prose", "writing metrics", "readability analysis", "check my paper's clarity", "writing score", or "assess writing quality". Produces a structured report with A-F grade, dimension scores, and prioritized fix suggestions.less

Write experimental details sections for ML papers with GitHub repository integration. Use when asked to "write experiments section", "document experim...more

Write experimental details sections for ML papers with GitHub repository integration. Use when asked to "write experiments section", "document experimental setup", "describe methodology", "write reproducibility details", or "experimental details". Extracts information from code to ensure accuracy.less

Visualize the logical skeleton of a paper's argument as a claim-evidence DAG. Use when asked to "map my argument", "does my logic hold", "argument str...more

Visualize the logical skeleton of a paper's argument as a claim-evidence DAG. Use when asked to "map my argument", "does my logic hold", "argument structure", "find logical gaps", or "why doesn't my paper flow". Flags missing links, orphan claims, circular reasoning, and unsupported assertions.less

Use this agent to evaluate papers, presentations, posters, or communications for target audience alignment. Impersonates different reader personas (re...more

Use this agent to evaluate papers, presentations, posters, or communications for target audience alignment. Impersonates different reader personas (reviewers, industry engineers, students, experts) to identify jargon, unclear explanations, and narrative gaps.less

Generate critical reviews of ML paper drafts simulating a skeptical reviewer. Use when asked to "review my paper", "find weaknesses", "critique this d...more

Generate critical reviews of ML paper drafts simulating a skeptical reviewer. Use when asked to "review my paper", "find weaknesses", "critique this draft", "what would reviewers say", "audit my contributions", "check my limitations section", or "assess my submission". Audits contribution claims against the evidence and limitations against the categories reviewers check, then provides harsh but constructive feedback to strengthen the paper before submission.less

Fact-check references in ML paper drafts. Use when asked to "verify citations", "check references", "fact-check bibliography", "validate citations", o...more

Fact-check references in ML paper drafts. Use when asked to "verify citations", "check references", "fact-check bibliography", "validate citations", or "audit references". Verifies papers exist on arXiv, checks author names, years, and titles against actual publications.less

Build a symbol table and check notation consistency throughout a paper. Detects overloaded symbols, undefined notation, and convention violations. Hyb...more

Build a symbol table and check notation consistency throughout a paper. Detects overloaded symbols, undefined notation, and convention violations. Hybrid: script-based regex extraction + LLM semantic analysis. Trigger: "check notation", "notation consistency", "symbol table", "find notation issues", "verify notation".less

Verify experimental claims in ML papers against source code repositories. Use when asked to "verify experiments", "check claims against code", "fact-c...more

Verify experimental claims in ML papers against source code repositories. Use when asked to "verify experiments", "check claims against code", "fact-check results", "audit experiments", or "validate paper against repo". Cross-references paper statements with actual implementation.less

Build a DAG of theorem/lemma/proposition dependencies across the paper. Computes criticality scores, maps assumption flow, and detects orphan lemmas o...more

Build a DAG of theorem/lemma/proposition dependencies across the paper. Computes criticality scores, maps assumption flow, and detects orphan lemmas or circular dependencies. Trigger: "map theorem dependencies", "theorem DAG", "dependency graph", "trace assumptions".less

Deep verify ALL paper claims with systematic evidence hierarchy. Supports parallel mode via the parallel-audit orchestrator. Activates when asked to "...more

Deep verify ALL paper claims with systematic evidence hierarchy. Supports parallel mode via the parallel-audit orchestrator. Activates when asked to "audit claims", "verify claims", "check paper claims", "claim verification", "evidence check", "verify evidence", or "quick evidence scan". Includes Quick Mode for rapid brainstorming checks.less

Stress-test theorems by systematically exploring what happens when assumptions are dropped or weakened. Generates low-dimensional test cases and bound...more

Stress-test theorems by systematically exploring what happens when assumptions are dropped or weakened. Generates low-dimensional test cases and boundary conditions. Trigger: "find counterexample", "stress test theorem", "test assumptions", "break this theorem", "assumption necessity".less

Use this agent to challenge arguments, identify logical fallacies, and expose cognitive biases. Supports iterative refinement through constructive adv...more

Use this agent to challenge arguments, identify logical fallacies, and expose cognitive biases. Supports iterative refinement through constructive adversarial thinking. Invoke during brainstorming, hypothesis formation, or before committing to claims.less

Comprehensive pre-submission paper audit combining reviewer simulation, claim verification, clarity analysis, notation checking, statistical validatio...more

Comprehensive pre-submission paper audit combining reviewer simulation, claim verification, clarity analysis, notation checking, statistical validation, and audience alignment. Use when asked to "audit before submission", "pre-submission check", "is my paper ready", "self-review", or "submission readiness". Runs 6 diagnostic passes in parallel and produces a unified readiness report.less

Decompose proofs into logical steps, check each step follows from prior ones, identify assumption usage, and flag gaps or unjustified leaps. The theor...more

Decompose proofs into logical steps, check each step follows from prior ones, identify assumption usage, and flag gaps or unjustified leaps. The theoretical analogue of claim-auditor. Trigger: "audit proof", "check proof", "verify proof", "proof verification", "find proof gaps".less

Use this agent to verify statistical rigor in ML papers - p-values, confidence intervals, significance tests, effect sizes. Activates when asked to "v...more

Use this agent to verify statistical rigor in ML papers - p-values, confidence intervals, significance tests, effect sizes. Activates when asked to "validate statistics", "check statistical rigor", "verify p-values", "statistical validation", or "check significance".less

Opt-in, review-gated usage feedback for Research Agora skills (RFC-0001). Use when asked to "enable agora feedback", "share skill feedback", "show my ...more

Opt-in, review-gated usage feedback for Research Agora skills (RFC-0001). Use when asked to "enable agora feedback", "share skill feedback", "show my skill usage stats", "submit agora feedback", "report skill usage", "disable feedback capture", or "purge my feedback spool". Nothing is ever sent without the user reviewing the exact payload and explicitly confirming.less

Use this agent to prepare ML code/data/models for public release with comprehensive checklists. Activates when asked to "package artifacts", "prepare ...more

Use this agent to prepare ML code/data/models for public release with comprehensive checklists. Activates when asked to "package artifacts", "prepare release", "reproducibility checklist", "code release", or "prepare camera ready".less

Analyze and refactor Python codebases to remove dead code, eliminate duplication, and simplify complexity. Use when asked to "simplify code", "remove ...more

Analyze and refactor Python codebases to remove dead code, eliminate duplication, and simplify complexity. Use when asked to "simplify code", "remove dead code", "find duplicates", "refactor", "clean up codebase", or "reduce complexity".less

Sync ML experiment results to paper drafts. Use when asked to "update results", "sync experiments", "pull latest metrics", "update tables from code", ...more

Sync ML experiment results to paper drafts. Use when asked to "update results", "sync experiments", "pull latest metrics", "update tables from code", "experiment to paper", "refresh results", or "sync paper with repo". Extracts metrics from code/logs and updates paper sections.less

Make publication figures for ML papers, in TikZ or matplotlib. Use when asked to "create a figure", "make a plot", "visualize results", "draw a neural...more

Make publication figures for ML papers, in TikZ or matplotlib. Use when asked to "create a figure", "make a plot", "visualize results", "draw a neural network", "make a diagram in LaTeX", "TikZ flowchart", "architecture diagram", "publication figure", "style matplotlib", or "format figures for a conference". **tikz** draws diagrams whose content you describe; **plot** renders data you supply, read from real files rather than invented.less

Generate HTCondor submission files and wrapper scripts for ML research jobs. Use when asked to "create condor job", "submit to cluster", "write .sub f...more

Generate HTCondor submission files and wrapper scripts for ML research jobs. Use when asked to "create condor job", "submit to cluster", "write .sub file", "htcondor setup", "cluster job", or "batch submission". Supports GPU/CPU jobs, parameter sweeps, multi-seed experiments, and ablation studies.less

Build, debug and lint a LaTeX paper. Use when asked to "build the pdf", "recompile", "compile the paper", "why won't my paper compile", "the pdf is st...more

Build, debug and lint a LaTeX paper. Use when asked to "build the pdf", "recompile", "compile the paper", "why won't my paper compile", "the pdf is still the old version", "fix latex errors", "debug latex", "parse the log", "fix LaTeX", "make LaTeX consistent", "check LaTeX style", or "standardize notation". Three modes over one source tree: **build** compiles with latexmk and proves the PDF is fresh, **debug** reads the log that build produced and fixes what it reports, **lint** greps for house-style and notation violations.less

Keep a paper's equations and the code implementing them in agreement, via the latex-code-sync CLI. Use when asked to "set up latex-code-sync", "link e...more

Keep a paper's equations and the code implementing them in agreement, via the latex-code-sync CLI. Use when asked to "set up latex-code-sync", "link equations to code", "annotate functions with equations", "verify equations match code", "check the paper against the implementation", or "does my code match my math". Three modes of one workflow: **setup** bootstraps the package and CI, **annotate** links functions to equations with decorators, **verify** runs the checker and reports mismatches.less

Decode what reviewers actually want, then write the response. Use when asked to "triage reviews", "plan my revision", "what do reviewers really mean",...more

Decode what reviewers actually want, then write the response. Use when asked to "triage reviews", "plan my revision", "what do reviewers really mean", "prioritize reviewer comments", "write rebuttal", "respond to reviewers", "draft rebuttal", or "address reviewer comments". Two modes in the order you work: **triage** decodes reviewer subtext and ranks what to fix, **respond** writes the point-by-point reply with every quantitative claim sourced.less