Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Use when auditing an ACM CoNEXT submission for HotCRP readiness, covering the choice between the December and June cycles, the mandatory paper-registration step, the ACM acmart page budget (long ≤16 + ≤4 appendix; short ≤10 + ≤2 appendix), double-anonymous review, the reproducibility badge opt-in due at submission, PACMNET publication, and desk-reject triage before the cutoff.
.claude/skills/brycewang-stanford-conext-submission/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | 12% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 42% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 114% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 59% | 0% |
| case-10 | ✗→✓ | ▲ Improved | 39% | 0% |
Run this audit before uploading to the CoNEXT HotCRP site. CoNEXT research papers are published as journal articles in PACMNET (Proceedings of the ACM on Networking), so the submission is judged like a journal manuscript that also earns a conference slot. Every number below was read from the CoNEXT 2026 call and the CoNEXT 2025 CFP on 2026-07-09 via search renderings (see ../../resources/official-source-map.md); treat them as a one-cycle snapshot and reopen the live call first.
CoNEXT runs two submission cycles a year, and choosing the wrong one costs months:
an earlier notification (31 Mar 2026) and, if you earn a major revision, a spring resubmission path.
and the same rolling PACMNET volume.
Both feed one conference and one journal volume, so the decision is about readiness and calendar, not prestige — submit to the nearer honest deadline your evidence is actually ready for. See conext-workflow for the full backward plan from each cutoff.
CoNEXT separates paper registration from paper submission:
5 December 2025). Miss it and the system will not accept the PDF later.
Register with the real title and abstract, not a placeholder — the abstract drives reviewer bidding, and a registered abstract that diverges from the final paper quietly worsens your TPC match.
acmart template, unmodified. This is the ACM path — do not carry an IEEE two-columnhabit over from a sibling venue's kit.
of appendices.
of appendices.
read to accept the paper should live only in an appendix or the artifact (see conext-supplementary).
CoNEXT runs double-anonymous review: authors and reviewers are mutually hidden. Because networking papers lean on testbeds, deployments, traces, and operator relationships, the leak surface is wide:
bash# Mechanical pass on the submission PDF and any anonymized artifact archive pdfinfo paper.pdf | grep -Ei 'author|creator|producer' pdftotext paper.pdf - | grep -nEi 'github\.com/[a-z0-9-]+|gitlab|zenodo\.org/record|acknowledg|grant|our (testbed|deployment|university)' | head unzip -l artifact.zip | grep -Ei '\.git/|\.DS_Store|/home/|/Users/' | head grep -rniE 'university|@[a-z0-9.]+\.edu|internal hostname|AS[0-9]+ .*our' artifact/ --include='*.md' | head
The networking-specific leaks: a testbed or deployment named after your lab, a topology figure showing internal hostnames or an AS you operate, a trace whose capture point identifies your institution, a data-availability link pointing at a personal GitHub, and commit metadata inside a zipped repository. Re-host artifacts behind an anonymizing service before upload.
The optional ACM reproducibility badges are administered by CoNEXT's reproducibility committee, and the critical timing trap is here:
later.
other artifacts.
runnable artifact strengthen the review (see conext-reproducibility).
| Finding at audit time | Severity | Real fix | |---|---|---| | Main text over the page budget | Desk-reject-grade | Cut or move to the appendix; references are unlimited but do not absorb body text | | Template altered (margins, shrunk font, wrong acmart option) | Named desk-reject ground | Recompile clean, recover space editorially | | Identity leak in PDF, artifact, topology figure, or testbed name | Anonymity violation | Re-anonymize and re-host; scrub PDF metadata | | Abstract/paper not registered by the earlier date | No submission slot exists | Nothing fixes this post-deadline — calendar it now | | Wanted a badge but did not opt in before submission | Ineligible for badging | Opt in now, before the cutoff | | Same study under review at SIGCOMM/NSDI/IMC | Dual-submission exposure | Withdraw one venue; verify the current concurrent-submission wording |
institution and recent collaborators — a day early; late conflicts are the classic midnight failure.
acmart variant is required.any AI-disclosure rule — all cycle-volatile; none should be assumed stable.
text[CoNEXT submission status] ready / blocked / needs work [Cycle] December / June, cutoff date + time zone [Registration] title/abstract/authors/conflicts locked by the earlier deadline? yes/no [Format] pages used (body/appendix/refs), acmart compliance [Anonymity] clean / leaks: <where> [Reproducibility] badge opt-in done before submission? artifact link present? [Fix queue] <ordered, with owners and dates before the cutoff>
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-06 | pass→pass | 22,245 | 15,343 | -31% | 1 | 1 | 0% | 3,126 | 4,565 | +46% | 0 | 0 | — |
case-01 | fail→fail | 42,382 | 17,682 | -58% | 1 | 1 | 0% | 6,472 | 3,756 | -42% | 0 | 0 | — |
case-02 | fail→pass | 30,103 | 20,539 | -32% | 1 | 1 | 0% | 3,875 | 4,341 | +12% | 0 | 0 | — |
case-03 | fail→fail | 32,232 | 24,078 | -25% | 1 | 1 | 0% | 4,501 | 4,588 | +2% | 0 | 0 | — |
case-04 | pass→pass | 20,362 | 13,627 | -33% | 1 | 1 | 0% | 2,477 | 3,932 | +59% | 0 | 0 | — |
case-05 | pass→pass | 24,093 | 20,942 | -13% | 1 | 1 | 0% | 3,231 | 4,289 | +33% | 0 | 0 | — |
case-07 | fail→pass | 15,572 | 14,497 | -7% | 1 | 1 | 0% | 2,371 | 3,371 | +42% | 0 | 0 | — |
case-08 | fail→pass | 14,117 | 8,961 | -37% | 1 | 1 | 0% | 1,520 | 3,252 | +114% | 0 | 0 | — |
case-09 | fail→pass | 18,135 | 14,062 | -22% | 1 | 1 | 0% | 2,082 | 3,308 | +59% | 0 | 0 | — |
case-10 | fail→pass | 17,893 | 12,137 | -32% | 1 | 1 | 0% | 2,054 | 2,858 | +39% | 0 | 0 | — |
case-11 | fail→pass | 16,385 | 6,293 | -62% | 1 | 1 | 0% | 1,699 | 2,688 | +58% | 0 | 0 | — |
case-12 | pass→pass | 21,738 | 10,755 | -51% | 1 | 1 | 0% | 1,946 | 3,395 | +74% | 0 | 0 | — |
case-13 | pass→pass | 19,394 | 12,457 | -36% | 1 | 1 | 0% | 1,765 | 3,000 | +70% | 0 | 0 | — |
case-14 | pass→pass | 15,943 | 8,545 | -46% | 1 | 1 | 0% | 1,908 | 3,160 | +66% | 0 | 0 | — |
case-15 | fail→pass | 16,924 | 11,438 | -32% | 1 | 1 | 0% | 1,857 | 2,820 | +52% | 0 | 0 | — |
case-16 | pass→pass | 12,322 | 12,698 | +3% | 1 | 1 | 0% | 1,827 | 2,998 | +64% | 0 | 0 | — |
case-17 | pass→pass | 19,909 | 13,605 | -32% | 1 | 1 | 0% | 2,162 | 3,051 | +41% | 0 | 0 | — |
case-18 | pass→pass | 13,882 | 16,200 | +17% | 1 | 1 | 0% | 2,117 | 3,332 | +57% | 0 | 0 | — |
case-19 | fail→pass | 22,615 | 12,010 | -47% | 1 | 1 | 0% | 2,302 | 2,975 | +29% | 0 | 0 | — |
case-20 | fail→pass | 18,695 | 15,785 | -16% | 1 | 1 | 0% | 2,186 | 2,807 | +28% | 0 | 0 | — |
case-21 | pass→pass | 26,697 | 20,645 | -23% | 1 | 1 | 0% | 2,744 | 4,266 | +55% | 0 | 0 | — |
case-22 | pass→pass | 17,875 | 19,896 | +11% | 1 | 1 | 0% | 2,005 | 3,938 | +96% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted. The headline lift of +41 percentage points is the difference between those two pass rates over the 22 comparable cases.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.