research-save

Save Research Finding

Save a research finding to .ai-memory/research.jsonl.

Steps

1. Parse the input: $ARGUMENTS contains the finding text. If empty, ask the user what finding they want to save.

2. Auto-extract metadata from the finding text:

   • topic: Extract a 5-15 word noun phrase summarizing the finding. Should be scannable and include library/tool names if relevant.
   • tags: Extract 1-5 keywords (lowercase). Include library names, concepts, and key technical terms.
   • staleness: Classify as:
     • "stable" — language behavior, protocol specs, math properties (won't change)
     • "versioned" — library/framework-specific behavior (set version_anchored if a version is mentioned)
     • "volatile" — external service behavior, API responses, rate limits (may change anytime)
   • confidence: Default to 0.9 for explicit user-provided findings.

3. Check for overlapping entries: Read .ai-memory/research.jsonl and look for entries with:

   • 2+ matching tags AND similar topic (substring match)
   • If found, show the existing entry and ask: "Existing finding found on this topic. Supersede it? (y/n)"
   • If superseding, set "supersedes": "<existing-id>" in the new entry.

4. Generate the entry:

   {
     "id": "<8-char-random-hex>",
     "ts": "<ISO8601-timestamp>",
     "topic": "<extracted topic>",
     "tags": ["<extracted tags>"],
     "finding": "<the finding text from $ARGUMENTS>",
     "source_tool": "claude-code",
     "source_context": "Explicitly saved by user",
     "confidence": 0.9,
     "staleness": "<classified>",
     "supersedes": null,
     "version_anchored": null
   }
   

5. Append the JSON line to .ai-memory/research.jsonl.

6. Run sync to update tool-specific files:

   node -e "require('${CLAUDE_PLUGIN_ROOT}/scripts/sync-tools.js').syncAll(process.cwd())"
   

7. Report: Show the saved finding with its topic, tags, and staleness classification.