{
  "id": "page:docs-reference-repos-clawhub-mco0l-caveman-claw-research",
  "_kind": "Page",
  "_file": "wiki/docs/reference-repos/clawhub/mco0l/caveman-claw/research.md",
  "_cluster": "wiki",
  "attributes": {
    "nodeKind": "Page",
    "sourcePath": "docs/reference-repos/clawhub/mCo0L/caveman-claw/research.md",
    "sourceKind": "repo-docs",
    "title": "mCo0L/caveman-claw",
    "displayName": "mCo0L/caveman-claw",
    "slug": "docs/reference-repos/clawhub/mco0l/caveman-claw/research",
    "articlePath": "wiki/docs/reference-repos/clawhub/mCo0L/caveman-claw/research.md",
    "article": "\n# mCo0L/caveman-claw\n\n- **Archetype**: clawhub-plugin\n- **Stars**: 3\n- **Last pushed**: 2026-04-12\n- **License**: MIT\n- **Discovered**: 2026-04-12\n- **Language**: TypeScript\n- **Fork**: No\n\n## Summary\n\nToken cost reduction plugin for OpenClaw with two mechanisms: (1) Output mode -- instructs the AI to respond in terse, fragment-based style (same technical accuracy, fewer words), and (2) Auto-compression -- algorithmically trims older messages in conversation history before each request (strips articles, filler, synonyms shortened).\n\nActivates globally by default on every new session. Tracks token savings via `/cc stats`. Simple, minimal approach -- no LLM calls, no vector databases, no external services.\n\n## Assessment\n\nLOW extractable value. The concept is simple and already partially covered by babysitter's compression layer. The \"caveman mode\" (terse AI output) is a prompt engineering technique, not a plugin architecture pattern. The auto-compression of older messages (article stripping, filler removal) is a basic NLP technique. Tracked for completeness but low priority.\n\n**Extraction priority**: LOW\n\n# Extractable Value: mCo0L/caveman-claw\n\n## Processes\n\nNone.\n\n## Plugin Ideas\n\n### 1. Terse Mode Plugin\n- **Category**: Context & Memory\n- **install.md**: Installs a session-level toggle that instructs the AI to respond in terse, fragment-based style for token savings. Also algorithmically compresses older conversation history by stripping articles, filler words, and shortening synonyms. Tracks cumulative token savings.\n\n## Library Mapping\n\n| Extractable Process | Library Status | Action | Existing Path | Target Placement |\n|-------------------|----------------|--------|---------------|------------------|\n| N/A | N/A | No multi-step processes identified - terse mode is prompt engineering technique | - | N/A |\n\n## Plugin Marketplace Mapping\n\n| Plugin Idea | Marketplace Status | Action | Existing Plugin | Target Placement |\n|-------------|-------------------|--------|-----------------|------------------|\n| Terse Mode Toggle | POTENTIAL OVERLAP | Session-level terse response mode with conversation compression | babysitter compression system | plugins/a5c/marketplace/blueprints/terse-mode-toggle/ |\n\n## Implicit Procedural Knowledge\n\n- **Dual-layer token reduction**: Combining output mode (prompt engineering for terse responses) with input compression (algorithmic trimming of history) addresses both sides of the token cost equation. Simple but effective for cost-sensitive environments.\n- **Global-by-default activation**: The `globalByDefault` setting pattern -- plugin activates automatically on every new session unless explicitly disabled. Relevant for babysitter plugins that should be always-on.\n",
    "documents": []
  },
  "outgoingEdges": [],
  "incomingEdges": [
    {
      "from": "page:docs-reference-repos",
      "to": "page:docs-reference-repos-clawhub-mco0l-caveman-claw-research",
      "kind": "contains_page"
    }
  ]
}