CtrlK
BlogDocsLog inGet started
Tessl Logo

uinaf/autoreview

Run structured Codex/Claude autoreview closeout: choose the target, collect schema-validated findings, and rerun tests plus review until clean.

84

1.08x
Quality

89%

Does it follow best practices?

Impact

74%

1.08x

Average score across 4 eval scenarios

SecuritybySnyk

Passed

No known issues

Overview
Quality
Evals
Security
Files

criteria.jsonevals/scenario-3/

{
  "context": "Tests whether autoreview chooses the correct branch/PR target for a clean committed branch instead of running an empty dirty-local review, and whether it preserves the closeout proof loop.",
  "type": "weighted_checklist",
  "checklist": [
    {
      "name": "Branch target selected",
      "description": "The plan selects branch review against the PR base, such as `--mode branch --base origin/main`, rather than local/dirty mode",
      "max_score": 18
    },
    {
      "name": "PR base discovery command",
      "description": "The commands include discovering the actual PR base with `gh pr view --json baseRefName --jq .baseRefName` or an equivalent reliable source",
      "max_score": 12
    },
    {
      "name": "Autoreview helper used",
      "description": "The commands run the bundled autoreview helper directly rather than asking another model to perform a generic review",
      "max_score": 14
    },
    {
      "name": "No empty local review",
      "description": "The plan explicitly avoids dirty-local/local mode because the working tree is clean and the committed branch diff is the real target",
      "max_score": 14
    },
    {
      "name": "No push requirement",
      "description": "The plan does not push the branch solely to make autoreview possible",
      "max_score": 10
    },
    {
      "name": "Focused proof loop",
      "description": "The plan says accepted findings require focused tests or proof plus a rerun of autoreview until no accepted/actionable findings remain",
      "max_score": 14
    },
    {
      "name": "Reject advisory weakness",
      "description": "The plan treats review output as advisory and says findings must be verified before fixing",
      "max_score": 10
    },
    {
      "name": "Compact closeout shape",
      "description": "The output is a concise plan with the requested sections and no verbose review transcript",
      "max_score": 8
    }
  ]
}

SKILL.md

tile.json