efo report for 2026-05-08

[github-actions]

Repo: EBISPOT/efo | Report date: 2026-05-08 | Window: 2026-04-24 – 2026-05-08

Minimal but functional agent activity in the 14-day window. Copilot authored one merged PR fixing ontology unsatisfiability issues (PR #2686), and the Copilot PR reviewer reviewed one human-authored PR (PR #2687). No agent summons recorded. Five open issues with no agent involvement were surfaced as potential opportunities.

• Agent-touched items: 2 of 12 scanned
• Agent-authored PRs: 1 (merged)
• Agent summons: 0

[github-actions]

Qualitative Review — EFO, 2026-05-08

What's Working

Copilot-authored bug fix landed cleanly. PR #2686 was opened by Copilot on 2026-04-29 to fix 4 EquivalentClass axioms that were causing EFO to produce a large number of unsatisfiable classes (owl:Nothing) under the ELK reasoner. It was reviewed by zoependlington and merged on 2026-05-07 — a full week from open to merge, suggesting it went through normal review rather than a rubber-stamp. This is the most substantive agent contribution in the window.

Copilot PR reviewer added value on a human PR. PR #2687 (adding 3 new "response to" drug terms) received a substantive automated review from copilot-pull-request-reviewer[bot] that described the new classes and their context within the ontology. Two human curators (aleixpuigb and zoependlington) then reviewed and the PR was merged. The bot review appears to have served as a useful first-pass summarization.

What's Not Working

No agent summons at all. In 14 days across 12 scanned items, no human invoked an agent via @copilot, @claude, or similar. The team isn't yet treating agents as on-call collaborators for issues or discussion threads — agent involvement is limited to autonomous Copilot triggers (PR authoring, auto-review) rather than curators actively pulling agents into conversations.

Five open issues received no agent attention. None of the potential opportunity issues show any agent involvement, even where the ask is relatively bounded.

Recurring Patterns / Friction

• The agent activity that does land (PR #2686) is technically correct enough to get through review, suggesting the barrier isn't quality — agents just aren't being directed at issue-level work.
• The NTR (New Term Request) workflow (issue #2607, issue #2613) depends on upstream ontologies (e.g., ChEBI terms not yet existing), which is a blocker an agent cannot resolve on its own. This pattern is a recurring friction point even for human curators.

Potential Missed Opportunities

The dossier surfaces five open issues; three are worth noting, though none is a slam-dunk without deeper issue-body detail.

1. Issue #2644: Clean up terms with amount/level
The ask is to standardize naming conventions for quantitative terms generated inconsistently with either amount or level (causing GWAS matching issues). This is a bounded, search-driven task — find affected terms, apply a consistent rename rule — and the kind of systematic cleanup an agent could draft a PR for. The main risk is that OWL term renaming has downstream ID/annotation implications; the issue would need to clarify scope before an agent could safely proceed.

2. Issue #2684: EVA-EFO — additional imports 26/04
A request to import 158 specific terms from MONDO and HP, with IDs listed in the issue body. This is highly structured: a fixed list of IDs, a known source ontology, and an established import pattern in EFO. An agent that can parse the ID list and generate ROBOT import commands or OWL stubs would have a concrete deliverable. The framing ("Let me know if you need...") suggests the requester is waiting on a curator to act.

3. Issue #2626: Request for a new trait pattern for disease progression terms
The issue asks for a new trait template/pattern covering rate or other characteristics of disease progression. The dossier excerpt is thin on specifics, and pattern-level ontology design decisions typically require conceptual curation judgment. Flagged here for completeness, but confidence is low that an agent could produce a ready-to-merge PR without significant human direction first.