/interknow:recall — Knowledge Recall

Query the knowledge base for patterns relevant to a topic or question.

When to use

• Before starting a review, to see what patterns are known
• When investigating a bug, to check for similar past findings
• When a domain is detected, to surface domain-relevant knowledge

Steps

1. Check qmd availability

Test if qmd MCP server is available by calling qmd:status. If unavailable, fall back to file-based search.

2. Build query

Compose a semantic query from:

• The user's topic or question
• Project domain (from .claude/intersense.yaml if available)
• File paths or symbols mentioned

3. Search knowledge

Note: Knowledge entries have been converged into docs/solutions/. Search there as the primary source. config/knowledge/ is a legacy fallback.

Primary: docs/solutions/

With qmd: Use qmd:vsearch against the solutions directory:

qmd:vsearch query="<semantic query>" path="docs/solutions" top_k=10

Without qmd (fallback): Read docs/solutions/**/*.md files and filter by keyword matching. Exclude INDEX.md, critical-patterns.md, and search-surfaces.md.

Legacy fallback: config/knowledge/

If no results found in docs/solutions/, also check ${CLAUDE_PLUGIN_ROOT}/config/knowledge/*.md for any unmigrated entries.

4. Filter and rank

• Exclude archived entries (in docs/solutions/archive/ or config/knowledge/archive/)
• Prioritize entries with recent lastConfirmed dates
• Prioritize provenance: independent over provenance: primed
• Cap at top 5 results

5. Present results

For each matching entry, show:

• Title (from filename)
• Finding (first paragraph of body)
• Last confirmed date
• Provenance status

If no matches found: "No knowledge entries match this query."

Output format

## Knowledge Recall: {query summary}

### 1. {entry-title}
{finding description}
Last confirmed: {date} | Provenance: {independent|primed}

### 2. {entry-title}
...

---
{N} entries searched, {M} matches found.