Developer Resource

Claude Code Skills

AI-assisted development tools built into the Millennium Dawn repository

The Millennium Dawn repository includes a set of Claude Code skills — slash commands that automate common development tasks. They are available to anyone using Claude Code (opens in new tab) with the repository open.

Setup: Claude Code reads .claude/ automatically when you open the repo. No additional configuration is needed.


Quick Reference

SkillWhat it does
/validateRun all validation tools against changed or staged files
/standardize <file>Auto-reformat a focus, event, decision, or idea file to MD conventions
/new-focus <TAG>Scaffold a new focus tree file with stubs and localisation keys
/add-leader <TAG>Scaffold generals, field marshals, and admirals using the correct count formulas
/lifecycle-check <TAG>Audit a branch against the focus tree lifecycle checklist
/search-filter-check <file>Validate search_filters on every focus against the approved filter list
/content-review <file>Full content quality checklist (economic, political, visual, military, AI)
/review-branchQuick standards review of the entire branch diff
/audit <file>Deep multi-agent audit: simplification + performance + content
/fix-issue <number>Find and fix a GitHub issue, then open a PR
/changelogSummarize branch changes and update Changelog.txt

How the Review Skills Relate

There are three review skills with different scopes:

/review-branch — fast, inline, single-pass. Covers coding standards, performance patterns, logic correctness, localisation, and quick content design checks. Best for PR review gating. Runs in one pass without sub-agents.

/content-review — focused on design quality. Checks the full developer content review guide: economic balance, political guidelines, visual standards, military counts, AI rules, and miscellaneous checklist items. Best when a developer thinks their content is submission-ready.

/audit — deep, multi-agent. Spawns three parallel agents (simplification, performance, content review) and merges their findings. Can also apply fixes. Best for a pre-merge cleanup sprint or a thorough file-level analysis.

These are additive tiers, not alternatives — running all three gives the most complete picture.


Lifecycle Skills

For developers building new country content, the recommended flow is:

  1. /new-focus <TAG> — scaffold the focus tree file
  2. /add-leader <TAG> — scaffold generals and admirals once OOB is drafted
  3. /search-filter-check <file> — validate search filters before submitting
  4. /lifecycle-check <TAG> — audit which lifecycle items are done
  5. /content-review — full design quality pass
  6. /review-branch — coding standards pass
  7. /validate — run automated validation tools
  8. /changelog — update the changelog

Validation and Standardization

/validate [staged] [strict] runs the Python validation tools in tools/. The staged flag limits checks to staged files; strict fails on any error (useful in CI).

/standardize <file> runs tools/standardization/standardize.py against a specific file and reports what was reformatted. This handles property ordering, logging insertion, and removal of empty blocks automatically.


Under the Hood

Skills are Markdown files in .claude/skills/<name>/SKILL.md. Each file contains instructions that Claude Code follows when the skill is invoked. Reference docs used by the skills live in .claude/docs/.

To update a skill’s behavior, edit the relevant .SKILL.md file. To add a new skill, create a directory under .claude/skills/ with a SKILL.md inside.