← Explore all skills

EVIDENCE AND EVALUATION / Undominated.ai

Undominated · Serving-mode attribution

Check that an explicit claimed score belongs to its named model or serving mode. Use when a model headline may have inherited a mode-specific score.

“Check that an explicit claimed score belongs to its named model or serving mode. Use when a model headline may have inherited a mode-specific score.”

01 / THE REASONING

Why this made the selection.

  • Pairs a bounded workflow with an offline Python checker and explicitly synthetic example inputs.
  • Created for recurring evidence and comparison failures: missing facts remain unknown and every conclusion keeps its scope.

02 / THE REVIEW RECORD

What we actually inspected.

Source review has boundaries.
A clear record is more useful than a “safe” badge.

Material inspected

  • Original definition and MIT licence
  • Bundled checker source, input contract and synthetic examples

Our findings

  • First-party resource authored by Undominated.ai; this is our own verification record, not an independent endorsement.
  • The download preserves the complete source and supporting files. Its manifest hashes identify the exact bytes.

Not established by this review

  • Model compliance with these instructions on arbitrary tasks
  • Native integration in every agent host

The review applies to the material and revision named here. A newer upstream release can change its behavior.

03 / PUT IT TO WORK

Add a skill to your workflow.

Upstream setup instructions ↗

DOCUMENTED COMMAND

npx --yes skills@1.7.1 add https://github.com/Lenvanderhof/Undominated.ai/tree/a67bd9b86fca7455ed208403d9ea6f9fe847cd99/skills --skill undominated-serving-attribution --agent codex --copy

Copying does not execute this command. It may retrieve a newer version than the reviewed source.

  1. Use the pinned command above in the intended project. Replace /absolute/path/to/project with an existing absolute directory when that argument is present.
  2. Download the complete bundle from https://undominated.ai/resources/skills/undominated-serving-attribution/bundle.zip and extract it into a new directory.
  3. Place the extracted directory at .agents/skills/undominated-serving-attribution/ or your host's documented skills directory. Preserve SKILL.md, scripts, examples and licence files together.
  4. Run the documented synthetic example from the skill directory with Python 3.10 or later before using your own evidence.

Before you start

  • An Agent Skills compatible host
  • Python 3.10 or later for the optional checker

THE COMPLETE REVIEWED DEFINITION

Read it before you reuse it.

Original source bytes, with attribution.
Review the host-specific setup notes above.

---
name: undominated-serving-attribution
description: Check that an explicit claimed score belongs to its named model or serving mode. Use when a model headline may have inherited a mode-specific score.
license: MIT
metadata:
  author: Undominated.ai
  version: "1.0.1"
---

# Serving-mode attribution

Keep model and serving-mode observations separate. Compare the actual `claimedScore` with the score for `claimSubject`; the presence of some score is insufficient. A mode score needs a named mode, including when the main claim concerns the model. Equality of numbers does not establish shared provenance.

## Input contract

Required fields:

- `model`: exact non-empty model identity.
- `servingMode`: exact non-empty mode identity or `null` when unknown.
- `claimSubject`: `model` or `mode`.
- `modelScore`, `modeScore`, `claimedScore`: non-negative decimal strings or `null`; never JSON numbers.

All fields must be supplied explicitly. A missing selected score or unknown claimed score requires review. A mode claim or supplied mode score without a named mode also requires review; a model-only claim may leave both mode fields null. Decimal equality is exact: `"100"` and `"100.00"` are equal; an additional decimal tail is preserved. There is no automatic substitution from the other score field.

## Deliverable and limits

Return the selected `expectedScore`, supplied claim, identities and any discrepancy. A pass establishes only consistency of the supplied identity and numeric fields. It does not show that a benchmark scored the named object or that two observations share conditions. Inspect the underlying source before making that attribution.

`examples/missing-mode.json` requires review. The conflict example uses a mode score while the model score is absent.

## Run the local check

Resolve paths relative to this skill directory:

```sh
python3 scripts/check.py examples/synthetic.json
python3 scripts/check.py /absolute/path/to/your-input.json
```

The examples are **synthetic**, not market measurements. Python 3.10+; standard library only. The script reads one explicit UTF-8 JSON file, prints JSON and makes no network requests or file writes.

Exit codes: `0` consistency check passed, `1` review required, `2` invalid input. CLI argument and malformed-input errors return structured JSON; `--help` displays usage text. Unknown or duplicate JSON fields, non-JSON constants and invalid types are rejected. Files are limited to 1 MiB; decimal strings to 1000 characters; identity/field names to 512 characters, with no surrounding whitespace or ASCII control characters.

This source-only skill is **not included in undominated-check@0.4.0**. Keep `SKILL.md`, the checker, examples and MIT licence together. A pass validates the bounded supplied-input contract, not production suitability or permission to publish. The skill does not authorize external actions.

The download contains SKILL.md. Extract the whole bundle; the supporting files are required. Inspect the included MANIFEST.json for file hashes.

By Undominated.ai. Exact upstream source ↗ · Licence · Attribution

SHA-256 a9a3aa54fb5c35300f666220b490b3fb128c4ab510ecfff3066adae78e87eb2d

Read the applicable licence
MIT License

Copyright (c) 2026 Undominated.ai

Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all
copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
SOFTWARE.

04 / FOLLOW THE EVIDENCE

The source trail.

Our notes are separate from the original resource.
Check upstream before adopting a new version.

  1. https://github.com/Lenvanderhof/Undominated.ai/blob/a67bd9b86fca7455ed208403d9ea6f9fe847cd99/skills/undominated-serving-attribution/SKILL.md

    Supports: Workflow and input requirements, Stated limitations

  2. MIT licence ↗Checked

    https://github.com/Lenvanderhof/Undominated.ai/blob/a67bd9b86fca7455ed208403d9ea6f9fe847cd99/skills/undominated-serving-attribution/LICENSE

    Supports: Redistribution terms and attribution

  3. Complete source bundle ↗Checked

    https://undominated.ai/resources/skills/undominated-serving-attribution/bundle.zip

    Supports: Full local source, supporting files and integrity manifest

Evidence & Ask