Update to Caveman

{
  "name": "caveman-compress",
  "description": "Use when the user explicitly asks to compress a memory, instructions, or todo file into Caveman format while keeping a readable backup.",
  "included_files": [
    {
      "relative_path": "README.md",
      "size_in_bytes": 5675
    },
    {
      "relative_path": "SECURITY.md",
      "size_in_bytes": 1727
    },
    {
      "relative_path": "agents/openai.yaml",
      "size_in_bytes": 237
    },
    {
      "relative_path": "scripts/__init__.py",
      "size_in_bytes": 224
    },
    {
      "relative_path": "scripts/__main__.py",
      "size_in_bytes": 30
    },
    {
      "relative_path": "scripts/benchmark.py",
      "size_in_bytes": 2491
    },
    {
      "relative_path": "scripts/cli.py",
      "size_in_bytes": 2471
    },
    {
      "relative_path": "scripts/compress.py",
      "size_in_bytes": 33727
    },
    {
      "relative_path": "scripts/detect.py",
      "size_in_bytes": 5169
    },
    {
      "relative_path": "scripts/validate.py",
      "size_in_bytes": 15236
    }
  ],
  "skill_md_contents": "---\nname: caveman-compress\ndescription: Use when the user explicitly asks to compress a memory, instructions, or todo file into Caveman format while\n  keeping a readable backup.\n---\n\n# Caveman Compress\n\n## Purpose\n\nCompress natural language files (CLAUDE.md, todos, preferences) into caveman-speak to reduce input tokens. Compressed version overwrites original. Human-readable backup saved as `<filename>.original.md`, but NOT beside the source file — it lives in an out-of-tree data dir (`$XDG_DATA_HOME/caveman-compress/backups/<parent-dir-name>/`, or `%LOCALAPPDATA%\\caveman-compress\\backups\\<parent-dir-name>\\` on Windows) so skill auto-loaders don't re-ingest it as a live file.\n\n## Trigger\n\n`/caveman-compress <filepath>` or when user asks to compress a memory file.\n\n## Process\n\n1. The compression scripts live in `scripts/` (adjacent to this SKILL.md). If the path is not immediately available, search for `scripts/__main__.py` next to this SKILL.md.\n\n2. From the directory containing this SKILL.md, run:\n\npython3 -m scripts <absolute_filepath>\n\n3. The CLI will:\n- detect file type (no tokens)\n- call Claude to compress\n- validate output (no tokens)\n- if errors: cherry-pick fix with Claude (targeted fixes only, no recompression)\n- retry up to 2 times\n- if still failing after 2 retries: report error to user, leave original file untouched\n\n4. Return result to user\n\n## Compression Rules\n\n### Remove\n- Articles: a, an, the\n- Filler: just, really, basically, actually, simply, essentially, generally\n- Pleasantries: \"sure\", \"certainly\", \"of course\", \"happy to\", \"I'd recommend\"\n- Hedging: \"it might be worth\", \"you could consider\", \"it would be good to\"\n- Redundant phrasing: \"in order to\" → \"to\", \"make sure to\" → \"ensure\", \"the reason is because\" → \"because\"\n- Connective fluff: \"however\", \"furthermore\", \"additionally\", \"in addition\"\n\n### Preserve EXACTLY (never modify)\n- Code blocks (fenced ``` and indented)\n- Inline code (`backtick content`)\n- URLs and links (full URLs, markdown links)\n- File paths (`/src/components/...`, `./config.yaml`)\n- Commands (`npm install`, `git commit`, `docker build`)\n- Technical terms (library names, API names, protocols, algorithms)\n- Proper nouns (project names, people, companies)\n- Dates, version numbers, numeric values\n- Environment variables (`$HOME`, `NODE_ENV`)\n\n### Preserve Structure\n- All markdown headings (keep exact heading text, compress body below)\n- Bullet point hierarchy (keep nesting level)\n- Numbered lists (keep numbering)\n- Tables (compress cell text, keep structure)\n- Frontmatter/YAML headers in markdown files\n\n### Compress\n- Use short synonyms: \"big\" not \"extensive\", \"fix\" not \"implement a solution for\", \"use\" not \"utilize\"\n- Fragments OK: \"Run tests before commit\" not \"You should always run tests before committing\"\n- Drop \"you should\", \"make sure to\", \"remember to\" — just state the action\n- Merge redundant bullets that say the same thing differently\n- Keep one example where multiple examples show the same pattern\n\nCRITICAL RULE:\nAnything inside ``` ... ``` must be copied EXACTLY.\nDo not:\n- remove comments\n- remove spacing\n- reorder lines\n- shorten commands\n- simplify anything\n\nInline code (`...`) must be preserved EXACTLY.\nDo not modify anything inside backticks.\n\nIf file contains code blocks:\n- Treat code blocks as read-only regions\n- Only compress text outside them\n- Do not merge sections around code\n\n## Pattern\n\nOriginal:\n> You should always make sure to run the test suite before pushing any changes to the main branch. This is important because it helps catch bugs early and prevents broken builds from being deployed to production.\n\nCompressed:\n> Run tests before push to main. Catch bugs early, prevent broken prod deploys.\n\nOriginal:\n> The application uses a microservices architecture with the following components. The API gateway handles all incoming requests and routes them to the appropriate service. The authentication service is responsible for managing user sessions and JWT tokens.\n\nCompressed:\n> Microservices architecture. API gateway route all requests to services. Auth service manage user sessions + JWT tokens.\n\n## Boundaries\n\n- ONLY compress natural language files (.md, .txt, .typ, .typst, .tex, extensionless)\n- NEVER modify: .py, .js, .ts, .json, .yaml, .yml, .toml, .env, .lock, .css, .html, .xml, .sql, .sh\n- If file has mixed content (prose + code), compress ONLY the prose sections\n- If unsure whether something is code or prose, leave it unchanged\n- Original file is backed up as FILE.original.md before overwriting — in the out-of-tree backup data dir (see Purpose), not beside the source file\n- Never compress FILE.original.md (skip it)\n"
}