agentsclimarketplace

Plan submission

Skill ShaishavMaisuria/research-paper-lifecycle-skills/skills/plan-submission

Builds a backwards submission timeline from a conference deadline - days remaining (AoE-aware), abstract-registration offsets, author-list freeze, OpenReview/CMT/EasyChair/HotCRP/PCS account lead times and per-system steps, supplementary-material and code size limits, reciprocal-reviewing duties, and a dual-submission audit. Use when the user wants to plan a submission, asks how long until a deadline, mentions abstract registration, submission checklist, deadline countdown, "what do I need before the NeurIPS/ICML/CHI deadline", or asks what happens after submitting (rebuttal window, notification, camera-ready). Phase-aware - produces pre-submission, under-review, and camera-ready plans from venues/ profiles and flags overdue steps.From its SKILL.md

Install
npx -y skills add ShaishavMaisuria/research-paper-lifecycle-skills --skill plan-submission

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

4 things to look at

  • reads credentialsReads from 1 credential source: `CONTACT_EMAIL`.
  • 23 stars23 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
  • runs commandsInstructs the agent to run 2 commands, including `python3 scripts/fetch_page.py "<cfp_url from the profile>"` and 1 more.
  • fetches URLsInstructs the agent to fetch 1 URL, including <cfp_url from the profile>.

SKILL.md

6.9 KB, ~1.5k tokens by cl100k_base, as published. Nobody here has run it

Plan Submission

Turn "the deadline is in N weeks" into a dated, phase-aware milestone plan. Works backwards from the paper deadline (or forwards through rebuttal and camera-ready if the deadline already passed), anchored to a venues/conferences/<venue>-<year>.yml profile. Companion skills: parse-cfp (creates the profile), preflight-check (the T-3d desk-reject lint), write-rebuttal, prepare-camera-ready.

When to use

  • "Plan my SIGSPATIAL submission" / "what do I need to do before the deadline?"
  • "How many days until the NeurIPS deadline?" (AoE confusion included)
  • "When do I have to register the abstract?" / "when does the rebuttal start?"
  • The user just picked a venue (via select-venue) and asks "what now?"

Inputs

  • A venue + year (and ideally a track). The profile must exist at venues/conferences/<venue>-<year>.yml; if it does not, run parse-cfp first — never plan from remembered dates.
  • Today's date, and any user-specific constraints (vacations, co-author availability, experiments still running).
  • CONTACT_EMAIL env var if the live re-verification fetch is used.

Process

  1. Resolve the venue profile. Locate venues/conferences/<venue>-<year>.yml. Missing or from a previous cycle → route through parse-cfp to create/refresh it before planning. Check the profile's verified: block — if confidence is not verified-live or verified.date is old, treat every date as suspect.

  2. Re-verify the critical dates against the live CFP. Mandatory before the user relies on the plan. Run:

    python3 scripts/fetch_page.py "<cfp_url from the profile>"
    

    Confirm the abstract deadline, paper deadline, and timezone on the live page (deadlines slip mid-cycle — extensions are common, but so are silent corrections). On mismatch, update the profile first. If the page is JS-rendered or down, say so and mark the plan "dates unverified".

  3. Build the timeline. Run:

    python3 scripts/build_timeline.py venues/conferences/<venue>-<year>.yml \
        --track "<track>" --today YYYY-MM-DD
    

    The script is offline and deterministic: it detects the current phase (pre-submission / under-review / camera-ready), computes days remaining for every profile deadline, and emits dated milestones with standard backwards offsets (T-21d dual-submission audit and reciprocal-reviewing check, T-14d system accounts, T-10d author-list freeze, abstract registration at its real date, T-5d supplementary, T-3d preflight, T-2d draft upload, T-1d final upload). --camera-ready YYYY-MM-DD supplies a deadline the profile lists as null. --format json for machine use.

  4. Customize the milestones using references/milestone-playbook.md: what each milestone actually requires, why the offset is what it is, supplementary/code size limits, reciprocal-reviewing mechanics, and the AoE math. Adjust offsets for the user's constraints (e.g. co-author on leave at T-10d → freeze the author list earlier). Never move a HARD deadline; only prep milestones are adjustable.

  5. Add per-system steps from references/submission-systems.md for the profile's review.submission_system — account lead times (OpenReview profile moderation can take ~2 weeks), what the submission form demands, conflict-of-interest entry, and each system's signature gotcha (HotCRP's "ready for review" checkbox, PCS metadata lock, EasyChair co-author email typos, CMT COI lists).

  6. Run the dual-submission audit interactively. Quote the profile's review.dual_submission policy verbatim, then ask the user to inventory every overlapping manuscript by ANY co-author that is under review or planned elsewhere (including workshop versions and arXiv plans). Flag conflicts against the quoted policy; when the policy is ambiguous, say so and point at the CFP — do not adjudicate.

  7. Present the plan: phase, key-dates table with days remaining, milestone table (OVERDUE items first, called out explicitly), per-system steps, policy gates, and the re-verify warning. If the paper deadline has already passed, do NOT pretend otherwise — plan the next phase and say plainly which deadlines are gone.

Output

A dated submission plan in Markdown: current phase, key dates with days-remaining (timezone noted), backwards milestone schedule with status (in Nd / TODAY / OVERDUE / PASSED), submission-system checklist, policy gates (dual submission, LLM policy, reciprocal reviewing), and the closing notice: "Dates from <profile> (verified <date>). Re-verified against <cfp_url> on <today>. Deadlines change mid-cycle — check the CFP again before each milestone."

Guardrails

  • Never invent a date. A null in the profile stays unknown in the plan, with an instruction on where to find it — a guessed deadline is worse than none.
  • Always state the timezone next to every deadline. AoE is common, not universal (SIGSPATIAL uses Pacific Time); a wrong assumption costs a day.
  • Never submit, register, create accounts, or click anything in any submission system on the user's behalf. The plan tells the user what to do; the user does it.
  • Past deadlines are reported as passed, never silently dropped or rescheduled.
  • One page per fetch-script invocation; cache stays under .cache/ (gitignored); never commit fetched CFP text.

Bundled files

  • references/milestone-playbook.md — what every milestone requires, default offsets and their rationale, supplementary/code limits, reciprocal reviewing, dual-submission audit, AoE math.
  • references/submission-systems.md — per-system walkthroughs: OpenReview, CMT, EasyChair, HotCRP, PCS.
  • scripts/build_timeline.py — deterministic phase-aware timeline builder (run it; do not compute date arithmetic by hand).
  • scripts/fetch_page.py — polite single-page fetcher for live CFP re-verification (vendored from parse-cfp so this skill is self-contained).
  • scripts/venue_profile.py — shared profile loader (YAML-subset parser + family merge), vendored for self-containment.

What ships with it: 5 files

65.3 KB alongside SKILL.md, 3 of them executable

scripts/

Gives 0 of the 12 instructions most plan spec skills give in ~1.5k tokens

Counted across 1,360 of the 2,617 authors here whose files we hold, read 2026-09-06

  • Ask one question at a timein 73 of 1360
  • Write the spec using the templatein 22 of 1360
  • Ask clarifying questions if neededin 19 of 1360, across 18 files
  • Wait for user confirmation before proceedingin 19 of 1360
  • Save plans to the plans directoryin 17 of 1360, across 13 files
  • Check for product marketing context firstin 16 of 1360, across 5 files
  • Read the plan file completelyin 16 of 1360
  • Order tasks by dependencyin 16 of 1360
  • Gather context from the conversationin 15 of 1360, across 9 files
  • Explore the codebase instead of askingin 15 of 1360, across 13 files
  • Wait for explicit user approvalin 14 of 1360, across 13 files
  • Quiz the user on the breakdownin 13 of 1360, across 7 files

Said here and by no other author read

  • Locate the venue profile YAML file.
  • Re-verify critical dates against the live CFP.
  • Run the timeline build script.
  • Customize the milestones using the playbook.
  • Add per-system steps for the submission system.
  • Run the dual-submission audit interactively.

Grouped from the skills themselves: near-identical wordings counted once, and counted by distinct author, so one author publishing three of these counts once. Length counted with cl100k_base; the agent that loads this file may tokenize it differently.

Keep looking

Skills are one crate of 325,949. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.