Add autoresearch plugin: autonomous experiment loop for optimization research#12
Merged
Conversation
…research New plugin with: - autoresearch skill: setup + experiment loop with MAD-based confidence scoring - autoresearch-finalize skill: extract clean branches from experiment history - autoresearch-experimenter droid: mission worker for orchestrated optimization - /autoresearch command: quick entry point - autoresearch_helper.py: stdlib-only Python helper for JSONL state, scoring, evaluation Inspired by karpathy/autoresearch and pi-autoresearch. Works for any optimization target (ML training, test speed, bundle size, etc.) both standalone and as a mission worker skill. Co-authored-by: factory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>
…y slash commands Co-authored-by: factory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>
…ion guidance Co-authored-by: factory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>
Co-authored-by: factory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>
… loop Co-authored-by: factory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>
factory-nizar
previously approved these changes
Mar 28, 2026
SKILL.md: - P0: Add init subcommand call in setup, include JSONL in initial commit - P0: Reorder keep flow: log to JSONL before git commit (prevents entry loss) - P0: Replace git clean -fd with backup/restore pattern for state files - P1: Add concrete baseline commands in Step 4 - P1: Add --asi flag to all log invocation examples - P1: Handle existing branch on resume (checkout || checkout -b) - P2: Fix evaluate text (compare against best-kept, not baseline) - P2: Skip user confirmation in mission worker finalization - P2: Reference status subcommand in resume section - P2: Show --metrics flag in log examples autoresearch_helper.py: - P0: Replace > 0 metric filter with status-based filter (supports zero/negative metrics) - P1: Use find_baseline() consistently in compute_confidence - P1: Add init-before-log guard (exit 1 if no config) Co-authored-by: factory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>
factory-nizar
approved these changes
Mar 28, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What
New plugin: autoresearch — an autonomous experiment loop for optimization research, inspired by karpathy/autoresearch and pi-autoresearch.
Try an idea, measure it, keep what works, discard what doesn't, repeat. Works for any optimization target: ML training loss, test speed, bundle size, build time, etc.
What's included
skills/autoresearch/SKILL.mdskills/autoresearch/autoresearch_helper.pyHow it works
autoresearch.md,autoresearch.sh,autoresearch.jsonl)METRIC name=valueoutput, evaluate with MAD-based confidence scoring, keep or discard, repeatKey features
autoresearch.checks.shfor correctness validation (tests, types, lint)/enter-missionfor longer campaigns; works as a mission worker skillInstall
Then invoke via
/autoresearchor let the model activate it when the user asks to optimize something.