Skip to content

Duplicate and prior-art checks need a run, so eight dead targets are opened for every live one #75

Closed

Description

@wolfgang-aura

Impact

record_duplicate_search and the prior-art read both write into a run
directory, so a coordinator has to create a run before it can find out whether
anyone has already fixed the issue. Most targets fail exactly that check, so
most runs exist only to be dropped. Each one costs the coordinator a sequence
of commands whose output stays in its context for the rest of the session.

Evidence

Hunt 20260907T164341Z-1ca91a: 24 runs, 21 dropped, 3 filed. Drop reasons:

cause count
duplicate or prior art already covers the issue 14
no longer reproduces at current upstream 2
review found an unresolved gap 3
screen freshness / unanswered contributor claim 2

Only the 3 review drops needed an agent. The other 18 were decidable from
public GitHub state before the run existed, and 14 of those from duplicate and
prior-art evidence alone.

Next action

A per-issue pre-screen that takes OWNER/REPO#N and writes to the screens
directory rather than a run directory, running the same narrow duplicate search
and prior-art read the run stage runs. Discovery shortlists through it, and
init-run is reached only by issues that passed. Related: #53, #41.

Activity

  1. wolfgang-aura commented on Sep 10, 2026

    @wolfgang-aura
    OwnerAuthor

    mailman/prescreen.py runs the duplicate, prior-art and claim checks against an OWNER/REPO#N before init-run, and writes the verdict beside the repository screens. Covered by tests/test_prescreen.py. Verified on main at 59da1d6.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions