Skip to content
cccskills
FreePublished on GitHub

archive-experiments

Preserve old experiment scripts with a dated Git tag, a named snapshot of the code, and add comments linking related GitHub issues to the archived files and cleanup pull request.

OriginalShow the original description

Archive explicitly named legacy experiment files by publishing a dated tag and canonical issue breadcrumbs.

Works withGitHub
See how to install

Good for

  • Preserving a snapshot of old experiment scripts
  • Linking experiment issues to archived files
  • Updating archive notices after a tag changes

Files included (1)

  • SKILL.md3.4 KB

SKILL.md (original)

Review the instructions the agent will receive before you install.

Skill: Archive Legacy Experiments

Overview

Retire old experiment scripts without losing their history: put the code behind a dated archive tag and leave a canonical comment on the originating GitHub issues so future readers can find the last snapshot.

Prerequisites

  • Local checkout of marin with push access to origin.
  • gh CLI installed and authenticated for marin-community/marin.
  • List of experiment issues and their corresponding filenames under experiments/.
  • The PR number for the cleanup you are referencing; if the directing human has not given one, ask them before posting anything.

Guidelines for Humans

1. Snapshot the experiments behind an archive tag

  1. Identify the commit that should house the archived experiments (usually main after the cleanup). Record it with git rev-parse HEAD.
  2. Pick a date-based tag name in the form archive/YYYYMMDD.
  3. Create and push the tag:
    TAG=archive/20251114
    git tag "${TAG}" <commit-sha>
    git push origin "${TAG}"
    
  4. If an older tag (e.g., exp_cleaning/YYYYMMDD) described the same snapshot, delete it locally and remotely so only the archive/ tag remains.

2. Post the canonical archive comment on each issue

  1. Confirm the cleanup PR number (PR_NUM); if nobody has provided one yet, ping the human operator for it before continuing.
  2. For every experiment issue, build the file URL: https://github.com/marin-community/marin/tree/${TAG}/experiments/<filename>.
  3. Use the standardized note below, swapping in the filename-specific URL and PR reference:
    This experiment has been archived to reduce clutter and preserve velocity (see PR #<PR_NUM>). It is last available in the `archive/YYYYMMDD` tag at <URL>. Please open an issue if you need help unarchiving it.
    
  4. Post the comment with gh issue comment:
    ISSUE=102
    FILE=exp102_classifier_ablations.py
    PR_NUM=1999
    URL="https://github.com/marin-community/marin/tree/${TAG}/experiments/${FILE}"
    gh issue comment "${ISSUE}" \
      --body "This experiment has been archived to reduce clutter and preserve velocity (see PR #${PR_NUM}). It is last available in the \`${TAG}\` tag at ${URL}. Please open an issue if you need help unarchiving it."
    
  5. Confirm the comment renders correctly (PR reference should auto-link and the URL should jump straight to the file in the archive tag).

3. Track progress

  • Keep a checklist of issue numbers while you work to avoid double-posting or skipping an experiment.
  • When the batch is complete, list the affected issues in your agent handoff or PR description.

Rules for Agents

  • Do not delete experiment files; archival work only tags commits and leaves issue breadcrumbs.
  • Always reference the current archive/ tag name in both inline code and URLs.
  • If the archive tag changes after comments were posted, update the affected issues rather than stacking multiple archive notices.

Validation

  • git ls-remote --tags origin | rg "${TAG}$" to ensure the tag is published.
  • gh issue view <issue> --comments | rg "archived to reduce clutter" to confirm each issue carries the canonical message.
  • Spot-check a few URLs to verify they open the expected files inside the archive tag.

See Also

  • .agents/skills/organize-experiments/SKILL.md for related curation workflows.

Reviews

No reviews yet. Tell others what you thought.

Comments

No login needed. Share how it worked for you, ask a question, or note where it did not run.

No comments yet. Be the first to write one.

Comments are published right away. Do not include personal information such as email addresses or phone numbers. Up to 2 URLs.

0 / 1,000 characters

This form is protected by reCAPTCHA, and the Google Privacy Policy and Terms of Service apply.

More skills from this repository

ab-test-zephyr

FreeSummary

Compare Zephyr changes using the same prepared dataset and matching execution settings. Analyze Finelog metrics for each processing stage to distinguish CPU efficiency, elapsed time, and peak memory changes, then report improvements, regressions, or tradeoffs.

  • Compare Zephyr performance before and after a PR
  • Assess speed gains against additional CPU usage
  • Check peak memory and out-of-memory regressions
marin-community/marin4KUpdated Oct 20, 2026

add-dataset

FreeSummary

Register Hugging Face datasets for Marin by inspecting their structure without downloading the full dataset. Identify text fields, configurations, and data splits, then add a dataset module and validate the mapping with a sample.

  • Registering a Hugging Face dataset for Marin
  • Inspecting text fields without a full dataset download
  • Checking configurations and data splits
marin-community/marin4KUpdated Oct 20, 2026

add-grant

FreeSummary

Turn a specified GCP access request into a reviewable Pulumi change and pull request. Select the narrowest suitable role and encrypt personal email identities, covering resource permissions and Cloud Run viewer access through IAP, an access gateway.

  • Prepare a request for GCS bucket read access
  • Add a viewer to a Cloud Run service protected by IAP
  • Prepare service account impersonation permissions
marin-community/marin4KUpdated Oct 20, 2026

Add or update Pallas/Mosaic kernels, specialized computation routines for JAX. The skill covers reference implementations, output and gradient checks, public wrappers, fallback behavior, performance benchmarks, and hardware-specific tuning when requested.

  • Adding a Pallas kernel for a JAX computation
  • Checking outputs and gradients against a reference
  • Maintaining backend selection and fallback behavior
marin-community/marin4KUpdated Oct 20, 2026

babysit-zephyr

FreeSummary

Launch and continuously monitor a specified Zephyr pipeline on Iris when asked. It checks coordinator and worker health, tracks stage progress, identifies stalled or repeated failures, and requires explicit approval before stopping or restarting jobs.

  • Launch a specified Zephyr pipeline and check startup
  • Monitor stage progress and worker health over time
  • Investigate stalled stages or repeated worker failures
marin-community/marin4KUpdated Oct 20, 2026

change-grug

FreeSummary

Try model or training changes in a copied Grug or Grugformer experiment, document the differences, and bring successful changes back into the base template while removing superseded experiment code and checking behavior.

  • Try changes to attention or masking
  • Experiment with optimizers or training loops
  • Bring successful changes into the base template
marin-community/marin4KUpdated Oct 20, 2026

See all skills by marin-community

Report a problem with this skill