Skip to content

Advisory reviewer: examples pin a nonexistent @v1, and v0.2.1's pinned harnesses refuse current models #475

Open

Description

@renmengye

Found while enabling the advisory reviewer on a lab repo (2026-10-03).

1. examples/review.yml references a tag that doesn't exist

  • examples/review.yml uses outerloop-science/outerloop/.github/workflows/advisory-review-agent.yml@v1. There is no v1 tag; the tags are v0.1.0 … v0.2.1 and v0.3.0rc1. docs/install.md uses @main.
  • Copying the example makes every run fail at workflow resolution ("This run likely failed because of a workflow file issue"), which looks like a configuration error rather than a missing ref. At least one other lab repo copied the example and has never had a working reviewer.

2. v0.2.1's pinned harness versions reject current models (both rounds were posted as outerloop:round-skipped)

  • backend: claude with model: claude-opus-5-5 fails with API Error: 400 Claude Code 2.1.229 does not support this model; version 2.1.280 or newer is required.
  • backend: codex with model: gpt-5.6-terra worked for several rounds on 2026-10-03, then started refusing: RuntimeError: run outerloop harness upgrade codex. v0.2.1 installs @openai/codex@0.130.0 (advisory-review-agent.yml line 192).

Ask

  • Publish a moving v1 tag, or a stable release whose pins work with current models, and point examples/review.yml at it. Alternatively, change the example to @main until v1 exists.
  • Ideally the skip comment would name the pinned harness version and the minimum version required, so users can tell "upgrade the reusable workflow ref" apart from an API outage.

Workaround meanwhile: pin @main (the version-pin reader from v0.3.0rc1).

Activity

  1. renmengye commented on Oct 4, 2026

    @renmengye
    MemberAuthor

    Update: pinning the reusable workflow to main (b749775) doesn't fix part 2. It installs @openai/codex@0.160.0 via the pin reader, and the codex round still fails with agent review produced no verdict … run outerloop harness upgrade codex (gpt-5.6-terra, 2026-10-04 02:10 UTC). So main's codex pin is also behind what the model requires.

  2. renmengye commented on Oct 4, 2026

    @renmengye
    MemberAuthor

    Thanks. Status:

    1. @v1 in the example: fixed in Codex host check: npm's script launcher finds its own vendored host #476. The example now uses @main with a note to pin a release tag or SHA (and reviewer_ref) in production. A moving v1 tag is a release-policy decision; it fits best once 0.3.0 is final.
    2. Pinned harnesses vs current models:
      • Codex: the refusal on 2026-10-03 came from mixing refs. The workflow file at @v0.2.1 installed Codex 0.130.0, while the default reviewer_ref: main checkout pinned 0.160.0 and asked for an upgrade. On @main the workflow reads the pins from the same checkout, so they agree. Install Codex's code-mode host beside the pinned binary #473 (merged) adds Codex 0.160's code-mode host for native installs, and Codex host check: npm's script launcher finds its own vendored host #476 keeps npm-installed Codex, the workflow's path, from being refused.
      • Claude: main pins Claude Code 2.1.272, still below the 2.1.280 that claude-opus-5-5 requires. A pin bump with checksums follows.
    3. Skip comment naming the versions: agreed; I'll fold it into the Claude bump: the skip note will name the pinned harness version and the minimum the API asked for.
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

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions