fbabelle

fbabelle/PrettySeriousResearcher

Plugin Claude Code

Plugin marketplace listing 1 plugin: research-paper-skills.

fbabelle/PrettySeriousResearcher

Plugin Claude Code

A 16-skill research-paper pipeline: topic selection, algorithm design, experiments, writing, and submission — with effort/cost tracking, anti-hallucination citation and provenance gates, publication-quality visuals, and a pre-submission reviewer panel.

fbabelle/PrettySeriousResearcher

Skill Claude CodeCodex

Phase 2 (20%) — evaluate candidate algorithms (efficiency, limits, maturity), pass a solvability gate, then design/implement the choice with tests, seeding ablation variants. Use when building the method once the topic is set (AI/Finance).

fbabelle/PrettySeriousResearcher

Skill Claude CodeCodex

Code-review gate for research code (Phases 2–3) — a 4-step design review (solves / pattern-fit / tradeoffs / validity) plus test discipline (tests ship and pass). Use after writing or changing an algorithm, harness, or run code, or on a PR.

fbabelle/PrettySeriousResearcher

Skill Claude CodeCodex

Phase 3 (30%) — smoke-test paid services, design the ablation matrix, probe local/supplier rate limits, build throttled+resumable run plans, track token/cost accounting. Use when planning or running experiments/ablations (AI/Finance, local).

fbabelle/PrettySeriousResearcher

Skill Claude CodeCodex

Finance gate — enforce statistical honesty (deflated Sharpe, PBO, multiple-testing, walk-forward, after-cost) and data licensing/point-in-time/survivorship provenance; names leakage-safe backtest infra. Use for any finance-facing topic.

fbabelle/PrettySeriousResearcher

Skill Claude CodeCodex

Pre-submission gate — an adversarial multi-model reviewer panel scores the draft on the target venue's rubric vs a numeric bar and returns ranked weaknesses + drafted rebuttals. Never auto-decides accept/reject. Use before submitting.

fbabelle/PrettySeriousResearcher

Skill Claude CodeCodex

Orchestrate an AI/Finance research paper end-to-end across four phases (topic, design, experiments, writing) with session clock, budget gate, and reflection checkpoints. Use to start or resume 'the paper', or ask 'where are we / what's next'.

fbabelle/PrettySeriousResearcher

Skill Claude CodeCodex

Results-integrity gate — every number, table, and figure in the draft must trace to a run artifact in runs/; surfaces only unresolved/placeholder values. The results sibling of the citation gate. Use at results-capture and before submission.

fbabelle/PrettySeriousResearcher

Skill Claude CodeCodex

Verify that every citation resolves to a real, current-status source whose content supports the claim; cover canonical origins and current work; drop unverifiable support, re-cite corrected work. Use when adding citations or reviewing related work.

fbabelle/PrettySeriousResearcher

Skill Claude CodeCodex

Periodic self-check — pressure-test the topic, design, or implementation against alternatives via adversarial search, then propose an optimization or conclude it's optimal. Use at phase boundaries, every 15 active hrs, or before submission.

fbabelle/PrettySeriousResearcher

Skill Claude CodeCodex

Standing conventions for a research-paper project — layout, docs backups, changelog, drafts rotation, README/architecture upkeep, commit/PR rules, tests, memory.md policy. Consult when committing, scaffolding, or unsure of a convention.

fbabelle/PrettySeriousResearcher

Skill Claude CodeCodex

Post-packet tail — execute the submission (portal checklist, anonymization, supplement, arXiv, artifact release), then the decision branches; evidence-mapped rebuttals, revise-and-resubmit, camera-ready. Use when submitting or when reviews arrive.

fbabelle/PrettySeriousResearcher

Skill Claude CodeCodex

Phase 1 (30%) — run the two-direction interview, do a prior-art / novelty-gap scan, score topics, and produce a paper skeleton with research questions and hypotheses. Use at a paper's start or when reframing the topic (AI/Finance).

fbabelle/PrettySeriousResearcher

Skill Claude CodeCodex

Track and report effort (agent-active hours + tokens) and cost (USD), and run the budget-finalization playbook. Use when reporting time/budget, updating a plan (re-report cost), at the budget gate, or when asked about hours, tokens, or spend.

fbabelle/PrettySeriousResearcher

Skill Claude CodeCodex

Find, rank, and time conferences/journals to submit to — by topic/domain fit, rank, deadline feasibility vs the estimated finish date, and prep/fees. Use when planning submission, picking a target venue, or as a finish date comes into view.

fbabelle/PrettySeriousResearcher

Skill Claude CodeCodex

Produce and review publication-quality visuals (tables, charts, flowcharts, diagrams, pseudocode, listings) — code-driven, one shared style. Use when creating, fixing, or reviewing a figure/table, picking a tool, or auditing a PDF's figures.

fbabelle/PrettySeriousResearcher

Skill Claude CodeCodex

Phase 4 (20%) — write and revise the paper (sections, figures/tables, tone, citations, appendix), version the drafts, keep README/architecture/changelog current. Use when drafting or revising the manuscript after results exist (AI/Finance).