AntFleet

Disagreement · 599e8a42-anthropic-1

Exit-code table contradicts step 2's 'gated' filter assumption

mismatch
repo 6f7fc663·PR #14·reviewed 1 week ago

Primary finding

Exit-code table contradicts step 2's 'gated' filter assumption

lowbuglow
  • skills/huggingface-trending/SKILL.md:56-66
The HF list endpoints (/api/models, /api/datasets, /api/spaces) do not include a top-level `gated` boolean by default for every entry; that field commonly only appears on the per-item detail endpoint. The skill instructs the agent to filter on `gated: true` from the list payload extracted in step 1, but step 1 never lists `gated` among the extracted fields. The filter rule is effectively unreachable unless an extra detail fetch is added — a subtle correctness gap that will silently let gated drafts through.

Recommendation

Either add `gated` to the step-1 extraction list (and document a per-item fetch if needed) or remove the gated-filter rule to avoid promising filtering the code can't perform.

Counterpart finding

Permalink guidance for datasets/spaces is inconsistent with stated constraint to use full URLs

lowdocs-gaphigh
  • skills/huggingface-trending/SKILL.md:55
  • skills/huggingface-trending/SKILL.md:172
The instructions require using full, correct HF URLs, but the earlier bullet shows relative paths for datasets/spaces. This can lead to implementers copying the relative form, contradicting the constraint and producing broken links.

Recommendation

Update the permalink bullet to use full URLs for all resource types, e.g.: models: https://huggingface.co/{id}; datasets: https://huggingface.co/datasets/{id}; spaces: https://huggingface.co/spaces/{id}.

Why this didn't post

This finding didn't meet AntFleet's unanimous agreement threshold. Both frontier models review every PR independently; only findings they both flag with the same severity and category are posted to the PR. This one fell through.

read the methodology →