mathodology-whole-project
starsforks
How to install mathodology-whole-project?
sweetcornna/mathodology/mathodology-whole-project
$npx -y skills add sweetcornna/mathodology --skill mathodology-whole-projectUse when running, backing up, transferring, restoring, archiving, or globally installing the Mathodology skills-only repository and its award workflows.
SKILL.md
| 1 | # Mathodology Whole Project |
| 2 | |
| 3 | ## Purpose |
| 4 | |
| 5 | This is the top-level orchestration skill for the current Mathodology GitHub |
| 6 | tree: running the award workflow, backing up, transferring, restoring, and |
| 7 | installing the skills globally. |
| 8 | |
| 9 | The repository is intentionally skills-only. It preserves Mathodology as an |
| 10 | AI-coding knowledge pack, not as a runnable app checkout. |
| 11 | |
| 12 | For repository maintenance, boundary and whitelist checks, and deciding whether |
| 13 | a change belongs on this branch, load `mathodology-project-orientation`. |
| 14 | |
| 15 | ## Current Shape |
| 16 | |
| 17 | The retained repository surface is: |
| 18 | |
| 19 | - `.claude/skills/**`: project skills and Codex-style metadata. |
| 20 | - `.claude/agents/**`: Claude Code project subagents for award-level modeling workflows. |
| 21 | - `.claude/workflows/**`: Claude Code workflow templates. |
| 22 | - `AGENTS.md`: tool-neutral entrypoint. |
| 23 | - `README.md` (Chinese-first) and `README_en.md` (English): public project overview. |
| 24 | - `docs/SKILLS*.md`, `docs/INSTALL*.md`, `docs/WORKFLOWS*.md`, and `docs/BACKUP.md`: skill, install, workflow, and backup documentation. |
| 25 | - `LICENSE` and `.gitignore`. |
| 26 | |
| 27 | Do not expect app source, CI workflows, deployment config, generated contracts, |
| 28 | datasets, package manifests, lockfiles, or installers in this branch. For the |
| 29 | full retained-file and deletion policy, load `mathodology-project-orientation`. |
| 30 | |
| 31 | ## Skill Set |
| 32 | |
| 33 | Load these skills as needed: |
| 34 | |
| 35 | - `mathodology-project-orientation`: current layout, retained files, deletion policy, and repository boundary checks. |
| 36 | - `mathodology-award-gates`: award-run gate schemas (handoff/gate/scorecard/decision_memo), the severity ladder, judge-panel thresholds, iteration budgets, run layout, the blind seat protocol, and the figure/PDF QA scripts. |
| 37 | - `mathodology-agent-pipeline`: archived knowledge about the former Python agent pipeline and the reusable phase workflow pattern. |
| 38 | - `mathodology-gateway-api`: archived knowledge about the former Rust gateway and API. |
| 39 | - `mathodology-web-ui`: archived knowledge about the former Vue web UI. |
| 40 | - `mathodology-dev-test-release`: skills validation (`validate_repo.py`) and archived dev, test, deploy, packaging, and release guidance. |
| 41 | - `mathodology-skill-authoring`: adding or updating project skills and metadata. |
| 42 | |
| 43 | ## Runtime Modes |
| 44 | |
| 45 | Choose the orchestration mode from the agent environment: |
| 46 | |
| 47 | - Claude Code project checkout: use `.claude/workflows/mathodology-award-submission.md` and dispatch the `.claude/agents/mathodology-*.md` subagents. |
| 48 | - Claude Code contest variants: when the contest is M3, HiMCM/MidMCM, IMMC/IM2C, leaderboard/data-science, operations/policy/business-case, or short-sprint style, also load `.claude/workflows/mathodology-contest-variants.md` and apply the matching adapter. |
| 49 | - Claude Code global skill install: load this skill and follow `docs/WORKFLOWS.md`; copy `.claude/agents/` and `.claude/workflows/` into the project if native project subagents are needed. |
| 50 | - Codex global skill install: run the workflow in multi-agents mode, dispatching independent agents for each phase, synthesizing their findings, then gating with a critic. In Phase 0, classify the contest type and apply the matching adapter from `docs/WORKFLOWS.md`. Treat the run as phase-sized and resumable; ask the user only for contest-critical details, then continue automatically when the gate passes. |
| 51 | |
| 52 | Codex start prompt: |
| 53 | |
| 54 | ```text |
| 55 | Use $mathodology-whole-project. Run the Mathodology 9-phase award submission workflow in Codex multi-agents mode. Work phase by phase: dispatch independent agents for analysis, modeling, evidence, coding, critique, and writing where applicable; synthesize their output; require result-density maps, figure/table sufficiency gates, and rendered-PDF figure QA; run the phase gate; then continue automatically. Pause to ask the user only for contest-critical details that would change requirements, data access, model choice, compute budget, or final submission constraints. For ordinary ambiguity, make a conservative assumption, record it in the phase log, and keep going. |
| 56 | ``` |
| 57 | |
| 58 | ## Codex Resumable Execution |
| 59 | |
| 60 | Codex may need multiple responses to finish the full award workflow. The lead agent should keep the run moving without asking the user to approve routine steps. |
| 61 | |
| 62 | Ask the user only when the answer would materially change: |
| 63 | |
| 64 | - official contest requirements, deliverable format, page limits, AI-use rules, or deadline |
| 65 | - access to private files, datasets, paid sources, credentials, or external services |
| 66 | - choice between plausible model routes with different scoring or feasibility risks |
| 67 | - compute, runtime, language, tool, or reproducibility constraints |
| 68 | - final claim, recommendation, or submission package decision that cannot be safely inferred |
| 69 | |
| 70 | For all other ambiguity, choose the safest defensible default, mark it as an assumption, and continue to the next task or |
People who liked this also viewed
6 picksType
- hyperframes-cliUse the HyperFrames CLI development loop: init, add, catalog, capture, lint, check, snapshot, compare, grade-compare, preview, play, present, beats, keyframes, single or batch render, publish, cloud, SkillsJul 2026
- request-refactor-planSkills for Real Engineers. Straight from my .agents directory.SkillsJul 2026
- analyze-projectRigor Analyze / Rigor Audit read-only skill for deep learning research repositories. Use when the user wants to read and understand a repository, inspect model structure and training or inference entrSkillsMay 2026
- explore-codeRigor Improve implementation leaf skill for auditable candidate implementation in deep learning research repositories. Use when the researcher explicitly authorizes exploratory work on an isolated braSkillsMay 2026
- ubiquitous-languageSkills for Real Engineers. Straight from my .agents directory.SkillsJul 2026
- scaffold-exercisesSkills for Real Engineers. Straight from my .agents directory.SkillsJul 2026
