docforge-llmslisted

# `docforge-llms` — sub-skill of `docforge` Canonical Markdown → AI-native projections (llms.txt, AGENTS.md, CLAUDE.md, MCP, .cursor/rules, .github/copilot-instructions, per-page .md). **One canonical. Many projections. Audience declared, or refused.** ## Layout ``` docforge-llms/ ├── SKILL.md # Body. ≤500 lines. Lazy-loads references. ├── README.md # This file. ├── references/ # Progressive-disclosure deepdives (load on demand). │ ├── llms-txt-spec.md │ ├── claude-md-anthropic.md │ ├── agents-md-spec.md │ ├── mcp-spec-essentials.md │ ├── cache-economics.md │ ├── chunkability-rules.md │ └── dual-output-pattern.md ├── templates/ # Output templates with deterministic placeholders. │ ├── llms-txt.template │ ├── llms-full-segmented.template │ ├── claude-md.template │ ├── agents-md.template │ ├── cursor-rules.template.mdc │ ├── copilot-instructions.template.md │ └── mcp-server.template.py ├── scripts/ # Capability implementations. │ ├── project.py # canonical → per-page .md + llms.txt │ ├── forge_agents_md.py │ ├── forge_claude_md.py │ ├── mcp_scaffold.py │ ├── chunkability_lint.py │ ├── cache_optimize.py │ ├── diagram_alt_referencer.py │ └── lib/ │ ├── frontmatter.py │ ├── tone_transformer.py │ └── token_counter.py ├── audiences/ # Per-audience profiles (knobs for tone transformer). │ ├── llm-runtime.profile.y