[eslint-refiner] ESLint Refiner — Daily Report 2026-07-04

[github-actions[bot]]

Overview

Focus this run: first review of the newly-added 8th rule require-await-core-summary-write (shipped by maintainers since the 2026-07-03 run — it was a nextPriorities candidate we'd flagged). Filed 3 non-duplicate refinement issues, all grounded against the live actions/setup/js corpus.

Key metrics

Metric	Value
Rules in plugin	8 (was 7)
Rule reviewed	require-await-core-summary-write
Issues filed	3 / 3 (limit)
Duplicates skipped	0 (no prior issue on this rule)
Live summary.write() sites	15 — all awaited (corpus disciplined)
Live un-awaited bugs	0

What the rule does well

• Catches bare core.summary.write() and coreObj.summary.write(), including chained addRaw(...)/addHeading(...).write().
• Correctly ignores unrelated fs.summary.write() / db.summary.write() / stream.write().
• Gates the await suggestion on the innermost function actually being async (no syntax-error autofixes).

Refinements filed

3 issues (1 high, 2 medium)

1. [HIGH · FN] Local summary alias not resolved. isCoreLikeIdentifier (/^core/i) means const summary = core.summary; summary.write() is never flagged — getRootIdentifier returns "summary". This is a live idiom: route_slash_command.cjs:20, check_workflow_timestamp_api.cjs:414/435 (all currently awaited, so the risk is preventive but the pattern is real). Ask: resolve simple const/let/destructure aliases of core.summary.
2. [MED · FP] /^core/i over-matches. Any identifier starting with core (coreData, coreResults, coreference) + .summary.write() is flagged. No live false positive today; ask to either tighten to scope-resolved @actions/core bindings or document the breadth with a codifying test.
3. [MED · FN] Non-bare-call statement forms escape. Visitor only handles ExpressionStatement → direct CallExpression, missing && / ?: / comma-wrapped calls and .then()-without-await chains. No live sites; keep void exempt as deliberate-discard.

Corpus signal

Every one of the 15 summary.write() call sites is awaited — the codebase is disciplined, so all three issues are hardening / preventive, with issue #1 anchored to a live aliasing idiom in two files. No prior open issue targets this rule (searched titles + summary.write).

Prior open issues (re-verified, not re-filed)

setOutput 42680/42682, catch-family 42188/42189/42190/42915, the two require-async-entrypoint-catch issues from 2026-07-03, and parseInt/json 42015–42017/41962/41963 remain valid and open. create_issue cap (3/run) was spent on the new rule; no duplicates emitted.

Next actions

• Watch whether the alias-resolution refinement (rejig docs #1) lands; if so, re-scan for newly-caught un-awaited sites.
• Consider recommending maintainers de-duplicate the two no-unsafe-*-catch-error-property rules (drift risk).
• Begin scouting a genuinely new defect class beyond the error-handling + summary family (e.g. core.setFailed non-string param, un-awaited exec.exec/fs.promises).

References:

• §28697370037

Generated by 🤖 ESLint Refiner · 158.7 AIC · ⌖ 12.5 AIC · ⊞ 4.7K · ◷

• expires on Jul 4, 2026, 10:23 PM UTC-08:00