Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Use when planning an ACM CoNEXT campaign backward from a deadline, covering the choice between the December and June submission cycles, the mandatory registration step, the one-shot major-revision window, the reproducibility badge opt-in due at submission, PACMNET camera-ready, and presentation — all on CoNEXT's distinctive two-cycles-per-year calendar.
.claude/skills/brycewang-stanford-conext-workflow/SKILL.md| Model | Eval pass | Runs |
|---|---|---|
| gemini-3.6-flash | 91% | 11 |
| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 51% | 0% |
| case-02 | ✗→✓ | ▲ Improved | -19% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 11% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 30% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 66% | 0% |
Plan the whole campaign backward from the cutoff. CoNEXT's defining feature is two submission cycles a year — a December deadline and a June deadline — both feeding one annual program and one rolling PACMNET volume. The first decision is therefore not "when do I write?" but "which cycle am I targeting, and is my evidence ready for it?" Every date below is the CoNEXT 2026 snapshot read on 2026-07-09 (see ../../resources/official-source-map.md); reopen the live Important Dates page before committing.
text[December cycle] register 5 Dec 2025 -> submit 12 Dec 2025 -> early-reject Feb 2026 -> notify 31 Mar 2026 -> camera-ready 30 Apr 2026 -> June 2026 PACMNET issue (major revision path) revision due 29 May 2026 -> resubmit 5 Jun 2026 -> camera-ready 31 Jul 2026 [June cycle] submit 5 Jun 2026 -> (its own review round, notification, and later PACMNET issue)
campaign rushed into December scores worse than a complete one in June.
so the two cycles are two entry points into the same program, not two separate conferences.
textT minus 12-16 weeks Lock the contribution and the evaluation platform (testbed/deployment/trace). Run conext-topic-selection: CoNEXT vs SIGCOMM/NSDI/IMC, and December vs June. T minus 8-12 weeks Build evidence with conext-experiments + conext-reproducibility. Pin traces, configs, firmware/OS versions NOW — they cannot be reconstructed later. T minus 4-8 weeks Draft against conext-writing-style and the worked example. Positioning with conext-related-work. Decide the body/appendix/artifact split (conext-supplementary). T minus 2 weeks Freeze the body. Build the anonymized artifact. Decide badging. T minus 1 week REGISTER (title/abstract/authors/conflicts). OPT IN for the reproducibility badge before the submission deadline if you want it. T minus 1-2 days conext-submission end to end: acmart budget, double-anonymous sweep, HotCRP fields, conflicts. T Submit; re-download and read the uploaded PDF cold.
text[Early-reject window] Some cycles notify clearly-below-bar papers partway through. No action but to plan the next cycle or a sibling venue if it comes. [Reviews / discussion] Two rounds + TPC meeting. Prepare an initial rebuttal if the cycle offers one (conext-author-response, Turn 1). [Decision] Accept / Reject / one-shot major revision. [One-shot revision] ~2 months to address the minimum-necessary changes; same reviewers re-read once; a shepherd checks the changes (conext-author-response, Turn 2). [Accept] Camera-ready to acmart + PACMNET metadata (conext-camera-ready). If you opted in, send the one-page artifact description within a week (conext-artifact-evaluation). [Reject] No appeal. Reroute via conext-topic-selection to a sibling venue or the other CoNEXT cycle, revised.
registration cutoff; a missed registration means no submission slot.
before submission and the one-page description after acceptance.
archive, and on the major-revision response letter.
required change to a tracked edit for the shepherd.
the one you are in.
fits ongoing work better (see conext-topic-selection).
text[Target] CoNEXT <year>, <December | June> cycle, submission <date + time zone> [Backward plan] <milestones with dates: platform freeze, evidence, draft, register, opt-in, submit> [Risk] <the milestone most likely to slip, and the fallback cycle/venue> [Post-submission] <decision-contingent branches: revision window, camera-ready, badge>
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 22,710 | 23,985 | +6% | 1 | 1 | 0% | 3,039 | 4,578 | +51% | 0 | 0 | — |
case-02 | fail→pass | 30,077 | 40,303 | +34% | 1 | 1 | 0% | 5,379 | 4,367 | -19% | 0 | 0 | — |
case-03 | fail→pass | 33,720 | 24,536 | -27% | 1 | 1 | 0% | 4,829 | 5,370 | +11% | 0 | 0 | — |
case-04 | pass→pass | 26,419 | 29,082 | +10% | 1 | 1 | 0% | 3,397 | 5,058 | +49% | 0 | 0 | — |
case-05 | pass→pass | 22,259 | 26,174 | +18% | 1 | 1 | 0% | 2,670 | 4,353 | +63% | 0 | 0 | — |
case-06 | pass→pass | 16,564 | 20,925 | +26% | 1 | 1 | 0% | 2,213 | 3,758 | +70% | 0 | 0 | — |
case-07 | fail→pass | 22,706 | 25,109 | +11% | 1 | 1 | 0% | 2,632 | 3,427 | +30% | 0 | 0 | — |
case-08 | fail→pass | 8,601 | 11,754 | +37% | 1 | 1 | 0% | 1,375 | 2,287 | +66% | 0 | 0 | — |
case-09 | fail→pass | 16,471 | 9,604 | -42% | 1 | 1 | 0% | 1,870 | 2,032 | +9% | 0 | 0 | — |
case-10 | fail→pass | 11,251 | 8,719 | -23% | 1 | 1 | 0% | 896 | 1,807 | +102% | 0 | 0 | — |
case-11 | pass→pass | 10,970 | 9,984 | -9% | 1 | 1 | 0% | 951 | 1,708 | +80% | 0 | 0 | — |
case-12 | fail→pass | 14,575 | 12,124 | -17% | 1 | 1 | 0% | 2,265 | 2,421 | +7% | 0 | 0 | — |
case-13 | fail→pass | 20,454 | 10,149 | -50% | 1 | 1 | 0% | 2,302 | 2,124 | -8% | 0 | 0 | — |
case-14 | pass→pass | 16,666 | 7,668 | -54% | 1 | 1 | 0% | 1,551 | 1,715 | +11% | 0 | 0 | — |
case-15 | fail→pass | 33,727 | 7,226 | -79% | 1 | 1 | 0% | 1,745 | 1,479 | -15% | 0 | 0 | — |
case-16 | fail→pass | 13,425 | 9,429 | -30% | 1 | 1 | 0% | 1,361 | 1,636 | +20% | 0 | 0 | — |
case-17 | fail→pass | 15,484 | 7,990 | -48% | 1 | 1 | 0% | 1,508 | 1,670 | +11% | 0 | 0 | — |
case-18 | fail→pass | 20,458 | 14,041 | -31% | 1 | 1 | 0% | 2,415 | 2,888 | +20% | 0 | 0 | — |
case-19 | pass→pass | 13,137 | 11,146 | -15% | 1 | 1 | 0% | 1,901 | 2,402 | +26% | 0 | 0 | — |
case-20 | fail→pass | 22,200 | 19,815 | -11% | 1 | 1 | 0% | 2,417 | 3,123 | +29% | 0 | 0 | — |
case-21 | pass→pass | 24,686 | 14,990 | -39% | 1 | 1 | 0% | 2,831 | 3,291 | +16% | 0 | 0 | — |
case-22 | fail→pass | 13,655 | 7,523 | -45% | 1 | 1 | 0% | 1,337 | 1,567 | +17% | 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 +68 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.