uberon report for 2026-05-13

[github-actions]

Report date: 2026-05-13 | Lookback: 14 days (from 2026-04-29)

Activity summary: 4 agent-touched items, 7 agent summons/references, 2 merged PRs, 1 open PR (agent-authored), 1 closed issue.

What's working: ai4c-agent[bot] is actively reviewing PRs with structured checklists. dragon-ai-agent successfully implemented a new ontology term end-to-end — from issue request through PR authorship, revisions, and merge. The review→revise loop between ai4c-agent[bot] and dragon-ai-agent on PR #3621 resolved a real logical inconsistency before merge.

What's not working / friction: PR #3643, opened by Copilot in January 2026, remains open and appears stalled — no substantive activity in the window beyond an automated inference diff from github-actions.

Opportunity: Issue #3706 (NTR: cnidarian polyp stage) is a recent new-term request with no agent involvement detected and may be a candidate for agent-assisted term addition.

[github-actions]

Qualitative Review — uberon, 2026-05-13

What seems to be working

ai4c-agent[bot] as PR reviewer. The agent left structured, checklist-based reviews on two PRs during the window. On PR #3641 (eyelid definition update), it posted two rounds of review — including both a job-linked summary and a formal approval checklist — demonstrating consistent review formatting. The reviews are readable and actionable rather than generic.

dragon-ai-agent full-cycle term authorship. The sinus of lymph node workflow (PR #3621, resolving issue #3586) is the clearest positive signal in the window. dragon-ai-agent authored the PR in November 2025 in response to a dosumis summons, then responded to reviewer feedback in May 2026 — adding explicit is_a axioms when aleixpuigb asked via @dragon-ai-agent. A subsequent ai4c-agent[bot] review caught a logical inconsistency in the class hierarchy (anatomical space vs. anatomical conduit), dragon-ai-agent revised, ai4c-agent[bot] re-reviewed and approved, and the PR was merged. This multi-round agent-to-agent collaboration with human oversight is the intended pattern and it worked.

Copilot as early reviewer. On PR #3641, copilot-pull-request-reviewer[bot] flagged a semantic problem early (incorrect "zone of skin" reclassification), which led to the more careful revision tracked in PR #3643. The early catch was substantive.

---

What seems not to be working

PR #3643 is stalled. This Copilot-authored PR was opened in January 2026 to address the eyelid reclassification. In the 14-day window it received only an automated inference diff from github-actions — no human review, no agent follow-up, no merge activity. The PR has been open for ~4 months with no substantive progress. This is a recurring friction point for Copilot-authored PRs: Copilot opens the PR and notifies the reviewer, but lacks the ability to self-revise or push the conversation forward if no one engages. The result is a growing backlog of agent-initiated PRs that stall without human triage.

Long PR latency before agent involvement. PR #3621 was opened in November 2025 and sat dormant for ~6 months before ai4c-agent[bot] reviewed it in May 2026. The agent review and revision cycle was efficient once it started, but the delay suggests the trigger for agent review is inconsistent or event-driven rather than systematic.

---

Recurring failure mode

The dominant pattern across both Copilot-authored PRs and older dormant PRs is: agents open or engage with a PR, then go quiet, and the PR stalls without a human to re-engage them. Neither dragon-ai-agent nor ai4c-agent[bot] appear to self-initiate follow-up after a period of silence. Human re-summons (like aleixpuigb calling @dragon-ai-agent in May) are what restart the loop.

---

Potential missed opportunity

Issue #3706 — NTR: cnidarian polyp stage

This is a new-term request (NTR) filed by ANiknejad on 2026-05-11. The issue body requests a developmental stage term for cnidarian species, citing specific transcriptomic use cases. No agent involvement has been detected.

The issue follows the same pattern as issue #3586 (sinus of lymph node), which was resolved end-to-end by dragon-ai-agent after a maintainer summons. If the dossier includes enough term detail in the issue body (preferred label, synonyms, definition, parent class), this would be a bounded, concrete task suited to dragon-ai-agent. However, cnidarian developmental stages are a specialized taxonomic domain, and the confidence of the recommendation depends on whether the issue body provides the full OBO-format metadata needed to create a well-formed term without hallucinating relationships. The dossier excerpt shows the issue is recent and the body exists, but does not quote the full metadata block — a maintainer should verify that before summoning an agent.

Recommendation: If dosumis or another maintainer confirms the issue body contains sufficient term metadata, a @dragon-ai-agent summons with explicit term parameters (as was done for issue #3586) is a plausible next step.