Skip to main content

Flagship Scorecard

Per-rule numbers for the 10 Interlace flagship ESLint rules, measured on real open-source repositories. Generated from the ILB-Flagship bench suite — dated snapshots committed to benchmarks/results/ilb-flagship/.

Head-to-head (warm)
2/4 rules at or faster than the peer
Green-field rules
5/10 with no peer equivalent
Snapshot age
12 days ago

Latency (cold → warm) and findings count

One row per flagship rule, measured on a real OSS repository. Cold is eslint --no-cache; Warm is the same command after a prior cold run with a stable cache location. oxlint caches implicitly via file mtime + content hash. A dash means the stack wasn’t exercised for that rule (green-field rules have no peer equivalent; not every rule has an oxlint port yet).

RuleRepo⭐TierOurs coldOurs warmOurs findingsPeer coldPeer warmoxlint coldTry
import-next/no-cyclenext.js131KT150,621 ms757 ms1266,206 ms754 ms1,307 msdocs
pg/no-unsafe-querysupabase78KT143,096 ms1,452 ms0———docs
secure-coding/no-hardcoded-credentialsvercel-ai15KT210,057 ms734 ms97312,744 ms647 ms—docs
secure-coding/no-redos-vulnerable-regexlodash60KT1384 ms301 ms1421 ms339 ms—docs
mongodb-security/no-unsafe-querypayload35KT210,365 ms801 ms299———docs
jwt/no-algorithm-nonesupabase78KT141,893 ms1,383 ms0———docs
browser-security/no-postmessage-wildcard-originnext.js131KT147,615 ms727 ms2———docs
react-features/hooks-exhaustive-depsnext.js131KT149,400 ms735 ms14749,177 ms813 ms525 msdocs
react-a11y/alt-textshadcn-ui100KT1———12,267 ms833 ms170 msdocs
vercel-ai-security/no-unsafe-output-handlingvercel-ai15KT29,882 ms645 ms1———docs

Cache effectiveness (median across rules)

StackMedian coldMedian warmΔCache benefit
Ours (ESLint)41,893 ms735 ms41,158 ms98%
Peer (ESLint)12,744 ms754 ms11,990 ms94%
oxlint native (peer)525 ms500 ms25 ms5%

How to read this

  • Latency is single-shot today. For SLO-grade numbers we owe you median-of-N with confidence intervals (tracked on the roadmap).
  • Findings count is filtered by the rule’s own ID prefix — parser errors and unrelated rules are excluded.
  • A higher findings count is not automatically better. See the source markdown report (linked above) for the per-rule ours-only / theirs-only triage tables.
  • The 10 rules are the flagship list at .agent/flagship-rules.md. Each is type-unaware and locked by the workspace test at scripts/__tests__/ecosystem-presets.test.ts.

Re-run with npm run ilb:flagship at the repo root. Tracks 10 rules per snapshot.

Provenance

Generated
2026-11-06 17:08 UTC
ESLint
v9.39.4
oxlint
1.63.0
Node
v24.22.2

Source JSON: benchmarks/results/ilb-flagship/2026-11-06.json · Schema ilb-flagship/v2.