[eslint-refiner] ESLint Refiner Daily Report - 2026-09-30

[github-actions[bot]]

Overview

Daily refinement pass over eslint-factory's 63 custom ESLint rules (up from 62 — no-single-char-string-replace shipped since yesterday). This run's highest-value signal was a rule-count bump plus a genuinely new feature commit (#64354, the ledger configuration feature — ~10 new ledger_*.cjs files): both this run's leads came from grounding rules against that fresh code rather than re-treading previously-reviewed corpus.

Key metrics

• 63 rules registered (was 62 yesterday); 1 never-before-reviewed.
• 1 issue filed today, plus 1 theme retired (2nd expiry, no refile).
• 5 open eslint-factory issues after today's filing (previous 4 + today's).
• Git-log-based diffing remains dead: HEAD's last commit is the 13th consecutive all-insertions shallow-clone squash artifact (8,139 files, 0 deletions) — novelty is now tracked via rule-count and live-file grepping instead.

Issue filed today

no-empty-catch-block: rationale comments without a magic keyword still flagged. The rule's intentional-ignore matcher only recognizes a fixed phrase list (intentional, best-effort, non-fatal, ignore, fall-through, no-op, swallow...because). Two brand-new empty catch blocks in ledger_store.cjs (lines 744-746 and 776-778, from #64354) explain their rationale clearly ("must not turn a durable append into a reported failure") but use none of those words, while sibling catches in the same file dodge the warning only because they happen to say "Best-effort." Filed with concrete acceptance criteria for broadening the matcher.

Retired this run

no-string-fallback-for-non-string-message's alias-type-narrowing false positive (#62560) expired for the 2nd time with no upstream fix. Per the now-established precedent (4 other themes retired the same way: try-catch-rule-utils.ts, no-json-stringify-equality, require-invalid-date-check-before-compare), this theme is retired from the refiling rotation — mentioned here for visibility only.

Full grounding detail

no-single-char-string-replace (first-ever review, new 63rd rule). Flags .replace("_", x)-style calls with a single-character string pattern as first-occurrence-only, suggesting replaceAll() (suggestion, not autofix). Exhaustively grounded all 6 live non-test call sites in the corpus:

• azure_devops_work_items.cjs:81 (_ → -)
• interpolate_prompt.cjs:241 (. → \., regex-escaping)
• parse_mcp_gateway_log.cjs:646,733 (T → , ISO timestamp)
• parse_mcp_scripts_logs.cjs:214 (, → ``)
• create_pull_request_helpers.cjs:219 (# → \#)

Every site is provably single-occurrence by the string's own structure (an ISO timestamp has exactly one T; the regex capture group that produces issueRef excludes #; env-derived names have exactly one dot separator) — so the rule's suggestion is harmless in 100% of live cases today. Also checked and ruled out: custom non-String .replace() implementations (none exist in the corpus), bracket-notation bypass (already excluded, matches the rule's own tests), regex-literal false triggers (already excluded via typeof check), and no-substitution template-literal patterns (`_` instead of "_", which would escape the Literal-type check) — zero live occurrences of the last one, so not filed. Clean bill of health.

require-fs-close-sync (re-scan against new code). Checked all 10+ new fs.openSync/fs.closeSync pairs introduced across ledger_store.cjs, push_ledger_changes.cjs, and ledger_append.cjs. Every open has a same-function close, including ones nested in their own try/catch inside a finally block — correctly handled since the rule doesn't treat try/catch/finally as separate function frames. No gap found.

Chronic/retired themes (unchanged, not re-checked this run): no-unsafe-promise-catch-error-property blanket-guard theory, bare RegExp(...) without new, no-json-stringify-equality indirect-variable gap, core-method-resolve.ts 2-hop alias theory, no-json-stringify-set-or-map parameter/return-binding gap — all previously retired after 4-5+ consecutive zero-grounding checks; not revisited today in favor of the fresher ledger-feature signal.

Known housekeeping gap (unfixable from this workflow): at least 2 confirmed cases (#61044, #61543) of a rule defect being fixed upstream in source without the tracking issue ever being closed — this workflow has no close_issue/add_comment safe-output. Flagging again for a human or PR-capable workflow to add issue-closure verification.

Next actions

• Watch today's new issue for expiry/fix; if it expires unfixed, re-grep the whole corpus (not just the ledger files) for the same "must not <consequence>" comment shape before refiling.
• Next run: check rule count first (cheap, reliable novelty signal); if unchanged, fall back to the rotating re-scan list (no-caught-error-interpolation, prefer-structured-clone, require-lastindex-reset-before-global-exec-loop, last touched 2026-09-11).
• no-single-char-string-replace is not yet documented in eslint-factory/README.md's rule table — low-priority doc debt, not worth a standalone issue yet.

Generated by 🤖 ESLint Refiner · claude · agent · 285.9 AIC · ⌖ 5.67 AIC · ⊞ 4.8K · ◷

• expires on Sep 30, 2026, 9:37 PM UTC-08:00

[github-actions[bot]]

This discussion was automatically closed because it expired on 2026-10-01T05:37:49.591Z.

Closed by Workflow