merge odds View on GitHub

CONTRIBUTION POLICY, BEFORE CODE

Read the rules before you touch the repo.

Merge Odds checks whether a project wants outside pull requests, whether it expects an issue first, and what it says about AI assisted work.

Check a project Install

8 measured projects. Every nondefault claim points to a source pinned at a commit.

POLICY FILE 001 READ FIRST
“We require a human in the loop at all times.”
directus/directus ai_policy.md at 197d141

THE READER

Policy first. Statistics second.

Select a measured project. A linked source means the project stated a rule that differs from the dataset default.

Showing directus/directus

CURRENT RECORD

directus/directus

MEASURED
01
External pull requestsAcceptedNo restriction found in files read
02
Issue firstNot requiredNo requirement found in files read
03
AI assisted codeConditionalRead ai_policy.md
04
AI written PR textDisallowedRead ai_policy.md
CASUAL ACCEPTANCE33%
MEDIAN TO MERGE4.09 days
P90 TO MERGE49.65 days
SAMPLE RECENCY51.45 days

This sample describes past pull requests from other people. It does not predict yours.

READ IN THIS ORDER

A percentage cannot overrule a maintainer.

The dataset leads with the project's own policy. Historical activity comes after it, with enough context to show where the sample can mislead.

Read the methodology
  1. 01

    Open the pull request template

    Restrictions can sit inside comments that disappear when GitHub renders the form.

  2. 02

    Check the pinned record

    Every quoted rule names the source file and commit where it was read.

  3. 03

    Read current policy files

    The dataset is a starting point. A project can change its rules after measurement.

  4. 04

    Use statistics as context

    Casual author activity says what happened in a sample. It grants no permission.

THE DATASET

Eight projects, one readable record each.

8 projects shown

Project External PRs Issue first AI code AI PR text Casual acceptance Median to merge Sources Measured
directus/directusyesnot requiredconditionaldisallowed33%4.09 days22026-08-07
n8n-io/n8nyesyesconditionaldisallowed47%0.57 days32026-08-07
payloadcms/payloadyesnot requirednot statednot statednot availablenot available02026-08-07
prisma/prismayesyesallowednot statednot availablenot available22026-08-07
strapi/strapiyesnot requirednot statednot stated41%5.99 days02026-08-07
supabase/supabaseyesyesnot statednot stated90%0.92 days12026-08-07
twentyhq/twentyyesnot requirednot statednot statednot availablenot available02026-08-07
withastro/astroyesnot requirednot statednot stated56%3.26 days02026-08-07

VET BEFORE WORK

Put the policy check in the workflow.

The vet-repo skill checks the pull request template first, reads current policy files, and reports statistics last. The vet-issue skill runs after a project passes, and confirms the defect still exists before any code is written.

CLAUDE CODE PLUGIN
/plugin marketplace add luantaraschi/merge-odds
/plugin install merge-odds
CLONE THE DATASET
git clone https://github.com/luantaraschi/merge-odds.git

WHAT IT CANNOT TELL YOU

No score predicts whether your pull request will merge.

Policy is measured at one commit. Statistics come from a limited sample. The weekly refresh verifies old quotes and recomputes activity, but it does not discover new policy language. Read the current repository before acting.