autoresearch
Skill2 starsChanged 14 days agoMITView on GitHub
Research to paper: develop a research question, compare venues, study matched accepted papers, plan and run experiments, then write and independently review an evidence-bound manuscript.
105 other projects use a version of this.
Adopted by
Most recently taken up by
- KPISTOLAS/claude-skillstheir own version, first seen Nov 16, 2026
- matt-riley/agent-skillstheir own version, first seen Nov 15, 2026
- tboome33/obsidian-mcp-routertheir own version, first seen Nov 15, 2026
- AndyMelnykov/obsidian-claudetheir own version, first seen Nov 15, 2026
- thisis-romar/opendeck-factorytheir own version, first seen Nov 15, 2026
- arbrain-ai/arbraintheir own version, first seen Nov 15, 2026
Used with
Projects with it also use aeon-update, auto-workflow and 10 more.
Other files people use alongside this one.
Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
aaronjmarsSkillaeon-update
Pull framework updates from the upstream Aeon repo into this instance - 3-way merges canon's new commits into a PR, never clobbering operator config.
Also in 4 projects that use this
aaronjmarsSkillauto-workflow
Two-mode aeon.yml workflow builder - analyze inspects URLs and emits a tiered, signal-verified skill-enablement plan plus an aeon.yml diff; enable flips slugs to enabled:true and opens a PR.
Also in 4 projects that use this
- +10 more from aaronjmars
Other versions105 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 105
- github/awesome-copilot★ 41ksame name
- Yeachan-Heo/oh-my-claudecode★ 41ksame name
- Yeachan-Heo/oh-my-claude-sisyphus★ 41ksame name
- Yeachan-Heo/oh-my-codex★ 34ksame name
- AgriciDaniel/claude-obsidian★ 16ksame name
- Orchestra-Research/AI-research-SKILLs★ 13ksame name
- zechenzhangAGI/AI-research-SKILLs★ 13ksame name
- Orchestra-Research/AI-Research-SKILLs★ 12ksame name
- advaitpaliwal/feynman★ 10ksame name
- companion-inc/feynman★ 10ksame name
95 more not listed
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository2 other files it gives its agents.
- AGENTS.mdAGENTS.md
- autoresearchSkill
The file
We split this skill into the rules it gives an agent
9 rules. 0 are common, 6 are distinctive, and 3 are very specific to this project.
0
Common
Most of the 101 edited versions we read keep these.
6
Distinctive
This author's own choices. Worth a look.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 23
Start with the current question, scope, angle, audience and available evidence.
- Distinctive: line 24
If a venue is unknown, compare actual candidates before freezing a target.
- Distinctive: line 41
Use the same state owner for rounds, snapshots, effects, bindings and recovery.
- Very specific: line 50
Read paper/references/budget.md for resource admission and recovery.
- Very specific: line 59
Use the stage's production_steps and paper/references/production.md for provided acquisition, recording, render/export, source variants and …
- Distinctive: line 66
Before promising an empirical run, preflight the actual condition matrix against the recording/statistics adapter.
- Very specific: line 81
Read ../../commands/autoresearch/<name>.md when the corresponding task arises.
- Distinctive: line 105
Use the files-only context and distinct Advisor assignments.
- Distinctive: line 109
No push, publication, deployment, paid call or expanded compute follows merely from loading the Skill.
“Common” is measured: kept by at least 6 in 10 of the 101 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Autoresearch Paper
The whole Skill carries a research project from materials and open questions to
verified evidence and a paper. Writing is the late write stage, after literature,
experimental or review evidence, and figure planning. Do not start with prose or
fabricate results to complete an outline.
Main entry
For /autoresearch, a research question, existing project or paper request, read
../../commands/autoresearch.md and follow the research workflow. The controller
reveals only the current stage's knowledge and executable handoff. No separate
paper subcommand is required.
Start with the current question, scope, angle, audience and available evidence.
If a venue is unknown, compare actual candidates before freezing a target. The
normal calibration sample is 8–15 accepted papers in the same direction, drawn
from the recent two editions or journal years; exceptions are explicit. Inspect
each measured figure/table and panel, not just caption counts. Venue knowledge
and the final obligation contract are ready before manuscript drafting.
Venue selection and the concrete plan budget are user-answer checkpoints.
Prepare the sourced candidates or budget proposal with ask-user; use
AskUserQuestion or the host's supported user-input interface (Codex can use
request_user_input_async when available), wait for an actual
answer, then apply its source-bound response. Show the budget default of no
implicit task cap and the optional calls, tokens, cost and total-time limits.
An increase requires another concrete question. Existing explicit answers may
be reused when they match the current question; execution delegation and a
preselected option do not supply an answer. See paper/references/user-input.md.
The stages are intake → venue selection/calibration → plan → literature → experiment/evidence synthesis → figures → write → independent review → delivery. Use the same state owner for rounds, snapshots, effects, bindings and recovery. A code keep, a completed tool or an overall score cannot satisfy a missing claim.
The only public resource-limit configuration is the Brief's budget.
Usage is receipt-backed in the same controller and survives host-session changes.
Thirty minutes prompts a checkpoint and continuation, not task termination.
Continue authorized internal repair; wait only on the affected external
dependencies. A stopped checkpoint requires an actual user stop or a verified
task-budget reason. Full completion still belongs to the original finalizer.
Read paper/references/budget.md for resource admission and recovery.
An active checkpoint preserves a continuation point. Keep working on the next
executable part of the original research request. Replies about venue, budget or
browser availability steer that same request; completing that intermediate step
does not finish it. Before ending a host turn, apply the continuation and handoff
conditions in the main command. Required user choices must actually be asked
through the host's user-input interface before they can justify waiting.
Use the stage's production_steps and paper/references/production.md for
provided acquisition, recording, render/export, source variants and review
assembly. The main CLI owns path resolution, round-request persistence and
export verification. Keep dataset/method choices in the research adapter;
do not copy a previous study's counts, filenames or reviewer roles into a run.
Before promising an empirical run, preflight the actual condition matrix against
the recording/statistics adapter. The supplied numerical adapter admits named
conditions with preregistered paired contrasts, averaging seeds within independent
cases. Preserve the full case × seed × condition denominator, including failed
and unstarted slots; operational failures block statistics. This descriptive
protocol supplies no universal statistical inference or scientific completion.
A two-arm development diagnostic does not satisfy a larger planned comparison.
For the separately bounded fixed-template vector profile, use
paper/references/local-response-protocol.md. Preserve its nested states,
source/receipt closure and two prospective freezes; its engineering tolerances
do not establish full-state identification or global stability.
Auxiliary protocols
These are tools for a bounded research subtask, not competing product entries.
Read ../../commands/autoresearch/<name>.md when the corresponding task arises.
| Command | Purpose |
/autoresearch:iterate | Original classic metric loop, including --classic, keep/discard and rollback |
/autoresearch:plan | Derive scope, metric and Verify for a bounded experimental/code subtask |
/autoresearch:debug, /autoresearch:fix | Reproduce and fix implementation failures |
/autoresearch:probe, /autoresearch:reason, /autoresearch:predict | Challenge assumptions and compare approaches |
/autoresearch:scenario, /autoresearch:improve | Examine alternatives and edge cases |
/autoresearch:learn, /autoresearch:evals, /autoresearch:regression | Record evidence, analyze progress and check regression |
/autoresearch:security, /autoresearch:ship | Explicitly requested engineering audit or release; external actions retain their authorization boundary |
For a predicate-bearing engineering orchestration task, read
references/engineering-orchestrator.md. --classic explicitly selects iterate;
--auto selects this auxiliary orchestration protocol. A bare natural-language
research goal always enters the research workflow.
Run bounded engineering work
The classic loop and orchestrator retain their existing limits, command screening,
hooks and rollback. references/fork-and-flush.md supplies fresh-context rounds
from the best commit: scheduled 33%/66%, a single 50% round below 12 iterations,
every 20 unbounded iterations, or five no-keep iterations replacing the next
scheduled round. Orchestrator plateaus permit two rounds without progress.
Use the files-only context and distinct Advisor assignments. Keep the authoritative
research ledger and evaluator outside the editable/rollback scope. Scientific
acceptance still requires the frozen contract and independent evidence review.
No push, publication, deployment, paid call or expanded compute follows merely from loading the Skill. Actual approved research resources and scope control work.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

