Spec Mint Core is a markdown-only Claude Code plugin (no build step, no dependencies) that replaces ephemeral AI coding plans with persistent, resumable specs. It also ships as a universal skill (SKILL.md) that works with Codex, Cursor, Windsurf, Cline, and Gemini CLI via npx skills add.
Version: 1.0.0 (plugin.json)
The plugin has two conceptual layers:
Plugin layer — commands/*.md (one file per slash command), agents/researcher.md (Opus-model deep research subagent), .claude-plugin/ (metadata). Claude Code reads these markdown files as behavioral instructions.
Data layer — .specs/ directory created in the consuming project root (not this repo). All tools share this directory. SPEC.md frontmatter is authoritative; registry.md is a denormalized index for quick lookups. See references/spec-format.md for the full format specification.
These files must stay in sync — changing one without the other will cause behavioral drift:
skills/specmint-core/SKILL.md is a symlink to ../../SKILL.md — never replace it with a real file.
CLAUDE.md,AGENTS.md, and.specs/are intentionally untracked in this repoAGENTS.mdprovides Codex-specific guidelines (seeSKILL.mdfor details)SKILL.mdmust work for all AI tools — the Claude Code Plugin section at the top is tool-specific and kept to ~20 lines- Spec format details (IDs, task codes, phase markers, sections) are in
references/spec-format.md— that is the single source of truth - Workflow details (forge phases, implement lifecycle) are in the respective
commands/*.mdfiles
- Edit
commands/*.mdto change slash command behavior - Edit
SKILL.mdto change universal skill behavior - Edit
references/spec-format.mdto change the SPEC.md format - Validate
.claude-plugin/*.jsonstays valid JSON after edits - Smoke-test changes:
claude plugin add /path/to/specmint-corein a disposable project, then run/forge,/resume, etc. - Windsurf users must replace the symlink at
.windsurf/skills/specmint-core/SKILL.mdwith a real file copy (Cascade doesn't follow symlinks)
Eval workspace at specmint-workspace/ (gitignored). Contains iterations 1-5 with evals covering forge workflow, resume, spec-then-stop, research depth, spec quality, implement, researcher spawn, acceptance criteria, and progress tracking. Eval definitions in specmint-workspace/evals/evals.json.
