methodic-research
Official@methodic-research
Methodic Research's Chronicle skills manage the full ML experiment lifecycle: hypothesis registration, variation authoring, training runs, checkpointing, distillation, and LaTeX paper authoring.
Agent Skills by methodic-research
Showing 40 vetted skills indexed across 1 GitHub repositories.
methodic-researchpaper-authoring
Draft LaTeX papers from Chronicle experiment records and attach compiled snapshots.
methodic-researchsynthesis
Proposes and queues next experiment variations grounded in Chronicle lineage, evaluation, and literature.
methodic-researchliterature-review
Synthesizes internal experiment records and external publications into a cited literature review.
methodic-researchevaluate-results
Evaluates Chronicle experiment runs against pre-registered hypotheses using real metrics.
methodic-researchchronicle-register-dataset
Registers existing gs:// or s3:// datasets with searchable physics metadata in Chronicle.
methodic-researchchronicle-delete-asset
Permanently deletes orphaned Chronicle assets not linked to any experiment.
methodic-researchchronicle-checkpoint-resume
Snapshots training environments and uploads resumable checkpoints to Chronicle runs.
methodic-researchchronicle-reproduce-arxiv
Registers an arXiv paper and imports its code repository into a Chronicle reproduction experiment.
methodic-researchchronicle-rebind-variation-git
Rebinds an open Chronicle variation from a code bundle to a pushed git ref.
methodic-researchchronicle-history-explorer
Explores Chronicle experiment history via semantic search, lineage DAGs, and retraction provenance.
methodic-researchfix-error-queue
Claims queued root causes, fixes bugs on branches, and opens GitHub PRs.
methodic-researchchronicle-move-experiment
Transfers a personal Chronicle experiment into an organization with optional team and visibility settings.
methodic-researchchronicle-import-reports
Imports research-report PDFs into a Chronicle organization's library with dedup and indexing.
methodic-researchchronicle-dataset
Upload, register, and load datasets as searchable assets linked to Chronicle experiments.
methodic-researchchronicle-fork-variation
Creates a new experiment variation branched from an existing committed variation's git ref.
methodic-researchchronicle-write-report
Attaches Markdown research write-ups with LaTeX math and figures to Chronicle experiments.
methodic-researchchronicle-publications
Registers and cites published works by DOI, BibTeX, or arXiv id in Chronicle experiments.
methodic-researchchronicle-bundle-variation
Packages external training code into a zip and registers it as a Chronicle variation code artifact.
methodic-researchchronicle-distill
Synthesize experiment findings across all variations into a review-gated report.
methodic-researchchronicle-propose-experiment
Creates a Chronicle experiment from a hypothesis with report, prompt, and optional commit.
methodic-researchchronicle-status
Inspects Chronicle experiment lifecycle state, run statuses, and retracted lineage ancestors.
methodic-researchchronicle-prep-variation
Creates a new Chronicle experiment variation with a cloned repo, agent branch, and config scaffold.
methodic-researchmethodic-feedback
Submits user feedback and gap reports to Chronicle's feedback endpoint and GitHub issues.
methodic-researchchronicle-tags
Tag assets and experiments with keywords and filter Chronicle searches by tag.
Frequently Asked Questions About methodic-research
FAQPage SchemaWhat tasks can I accomplish with methodic-research's Chronicle skills?▼
You can propose and commit experiments with pre-registered hypotheses, author and fork variations, execute training runs with W&B metric linking, checkpoint and resume crashed runs, register datasets from gs:// or s3://, survey literature, distill cross-variation findings, retract invalid results, and author LaTeX papers attached to experiments.
Who are these Chronicle skills designed for?▼
They target ML researchers and engineers running structured experiments, plus autonomous agents acting on their behalf. Personas include researchers reproducing arXiv papers, engineers launching SageMaker training jobs, and teams managing org-scoped report libraries, dataset catalogs, and experiment lineage with human approval gates.
How do Chronicle skills handle training execution and durability?▼
chronicle-run-variation drives the run lifecycle (start, heartbeats, succeed/fail) while your own training code executes. chronicle-checkpoint-resume captures environment snapshots and uploads resumable checkpoints; the sagemaker skill points checkpoints at /opt/ml/checkpoints for S3 sync and spot-interruption recovery with GCS cross-cloud fallback.
What are the prerequisites for using these skills?▼
You need access to a Chronicle organization and its managed experiment git repos. Literature search requires a configured literature MCP; metrics require W&B availability; dataset registration uses existing gs:// or s3:// URIs; SageMaker jobs need requirements.txt dependencies and the CHRONICLE_* environment variables injected by the launcher.
How does Chronicle enforce research integrity and human oversight?▼
Variations require pre-registered hypotheses with expected outcomes; distilled reports are review-gated until the experiment owner approves them. Retraction preserves lineage and audit trails while invalidating outputs, deletions require explicit confirmation, and the synthesis skill queues nothing without explicit human acceptance.