Conference Workshop Intel

Overview

Use this skill for public-source AI conference and workshop intelligence. The default closure works without private credentials: official conference sites, OpenReview public visibility, ai-workshop-tracker data, ACL Anthology, CVF Open Access, PMLR, NeurIPS proceedings, and source-specific public pages.

Decision Tree

1. Identify the target:
   • conference/year: NeurIPS 2025, ICML 2026, CVPR 2025, EMNLP 2025;
   • workshop ecosystem: workshop list, deadlines, accepted workshop papers, topic tags;
   • policy/CFP: call for papers, review policy, ethics policy, compute policy, reviewer guidance;
   • program signal: accepted papers, oral/spotlight, best paper, award, schedule;
   • OpenReview venue or forum URL;
   • cross-conference trend or monitoring task.
2. Run a route plan before crawling:

python skills\conference-workshop-intel\scripts\conference_workshop_intel.py plan `
  --target "NeurIPS 2025 workshops accepted papers CFP review policy" `
  --venue NeurIPS `
  --year 2025 `
  --needs workshops,accepted-papers,cfp,policy,awards,report

3. Execute only the needed lane:
   • Workshop list, deadlines, and OpenReview-hosted workshop accepted papers: use ai-workshop-tracker first, then official workshop pages.
   • ICLR/ICML/NeurIPS public submissions, decisions, reviews, and workshop venues: use OpenReview public pages or openreview-py.
   • ACL/EMNLP/NAACL/COLING proceedings: use ACL Anthology.
   • CVPR/ICCV/ECCV/WACV papers: use CVF Open Access and official conference program pages.
   • ICML/AISTATS/COLT papers: use PMLR plus official conference pages.
   • NeurIPS accepted papers: use NeurIPS proceedings and official neurips.cc pages; use OpenReview when the year/venue exposes the needed public fields.
   • AAAI/IJCAI/KDD/WWW/SIGIR: use official venue sites first, then DOI/open metadata enrichment.
   • Best paper/oral/spotlight/award/policy: use official program/blog/CFP pages first; third-party lists are discovery-only until verified.
4. Normalize before synthesis. Reports should cite normalized row IDs and source URLs, not loose browser state.

Full Local Setup

For the no-secret local setup and a public network smoke test:

powershell -ExecutionPolicy Bypass -File skills\conference-workshop-intel\scripts\setup_conference_workshop_intel.ps1 -RunNetworkSmoke

For a lightweight smoke without installing dependencies:

powershell -ExecutionPolicy Bypass -File skills\conference-workshop-intel\scripts\setup_conference_workshop_intel.ps1 -SkipPythonDeps -RunNetworkSmoke

This creates an isolated runtime under runtime/conference-workshop when the skill is installed globally or in the repo. It also creates a private .env placeholder without storing any secret. Read references/full-setup.md before configuring optional credentials.

External API Closure

Required credentials: none.

Use $external-api-onboarding before setting optional credentials such as OPENREVIEW_USERNAME, OPENREVIEW_PASSWORD, SEMANTIC_SCHOLAR_API_KEY, GITHUB_TOKEN, HF_TOKEN, or PAPER_SEARCH_MCP_*. Store them only in private user-level storage such as:

%USERPROFILE%\.codex\skills\conference-workshop-intel\.env

For visible provider-page assistance:

powershell -ExecutionPolicy Bypass -File skills\conference-workshop-intel\scripts\assist_conference_workshop_auth.ps1 -Provider openreview,semantic-scholar,github

The helper opens official provider pages only. The user completes login, MFA, CAPTCHA, key creation, and secret copy. Do not paste secrets into chat.

Source Routing

Read references/source-routing.md before selecting a backend.

• Use ai-workshop-tracker for ML workshop editions, deadlines, topic tags, OpenReview venue discovery, and accepted-paper caches.
• Use official venue sites for CFP, review policy, ethics policy, awards, oral/spotlight, schedules, and workshop pages.
• Use OpenReview for public submissions, decisions, reviews, scores, venue groups, and workshop venues when the venue exposes them.
• Use ACL Anthology, CVF Open Access, PMLR, and NeurIPS proceedings for official paper metadata and PDFs.
• Use deadline trackers, best-paper lists, and Paper Copilot-style pages only as secondary discovery seeds unless verified against official pages.

Common Commands

Classify a URL:

python skills\conference-workshop-intel\scripts\conference_workshop_intel.py inspect-url `
  "https://neurips.cc/virtual/2025/events/workshop"

Print the normalized artifact contract:

python skills\conference-workshop-intel\scripts\conference_workshop_intel.py schema

Generate a run scaffold:

python skills\conference-workshop-intel\scripts\conference_workshop_intel.py scaffold `
  --output-dir output\conference_workshop\neurips_2025_workshops `
  --target "NeurIPS 2025 workshop ecosystem" `
  --venue NeurIPS `
  --year 2025 `
  --needs workshops,deadlines,accepted-papers,cfp,policy,awards,report

Normalize a raw JSON/JSONL/CSV/YAML capture:

python skills\conference-workshop-intel\scripts\conference_workshop_intel.py normalize `
  --input output\conference_workshop\neurips_2025_workshops\raw\workshops.json `
  --source ai-workshop-tracker `
  --kind workshops `
  --output-dir output\conference_workshop\neurips_2025_workshops\normalized

Run a cheap public-source smoke:

python skills\conference-workshop-intel\scripts\conference_workshop_intel.py public-smoke

Output Contract

Use this directory shape for substantial runs:

• raw/: untouched API JSON, scraped HTML, YAML, CSV exports, screenshots, PDFs, and logs.
• normalized/events.jsonl: conference/year/event rows.
• normalized/workshops.jsonl: workshop editions, deadlines, websites, OpenReview venues, topics, and paper-list sources.
• normalized/papers.jsonl: accepted paper, submission, proceedings, oral/spotlight, and workshop-paper rows.
• normalized/policies.jsonl: CFP, review policy, ethics policy, compute policy, reviewer guidance, and policy-change rows.
• normalized/awards.jsonl: best paper, honorable mention, test-of-time, oral, spotlight, and award rows.
• normalized/artifacts.jsonl: local files, downloaded assets, screenshots, logs, and generated reports.
• sources.csv: source review table with URL, source type, priority, and status.
• manifest.json: plan, commands, timestamps, limits, credential policy, and blockers.
• reports/summary.md: human-facing synthesis grounded in normalized row IDs.

Read references/output-schema.md before merging multiple sources.

Guardrails

• Default work must remain useful without MCP, paid services, private tokens, cookies, proxies, or login-gated data.
• Prefer official venue pages, official proceedings, and OpenReview public visibility over search snippets and aggregators.
• Do not bypass paywalls, CAPTCHAs, private reviews, login gates, rate limits, robots restrictions, or license controls.
• Keep credentials, cookies, browser storage, proxies, headers, .env files, and provider account details out of git and final answers.
• Mark source priority as primary, secondary, archive, or fallback for every normalized row.
• Preserve source URL, source ID, fetched timestamp, venue, year, edition, and provenance fields.
• For policy and award claims, capture the exact official page URL and page date or crawl timestamp.
• Run a small public smoke before bulk crawl/download and record blocked or stale sources in manifest.json.

Resources

• scripts/conference_workshop_intel.py: planner, URL inspector, run scaffold generator, schema printer, lightweight normalizer, and public-source smoke test.
• scripts/setup_conference_workshop_intel.ps1: no-secret local runtime setup, optional ai-workshop-tracker clone, and public smoke tests.
• scripts/assist_conference_workshop_auth.ps1: visible provider-page opener and optional secret-storage command guidance.
• references/full-setup.md: setup matrix, optional credentials, provider pages, and smoke-test commands.
• references/source-routing.md: detailed source routes and fallback policy.
• references/output-schema.md: normalized fields and merge policy.