Search And Discovery Optimization

• Input: one or more pages plus their current titles, tags, browse paths, namespace placement, and local topology context
• Decision model: identify the weakest discovery link and recommend the smallest contract-aligned structural improvement
• Output: a discovery review that names the issue category, proposed structural fix, and any governed follow-up required
• Handoff rule: deterministic refreshes or topology changes route to existing wrappers and governance review rather than executing here

Assertions

• Discovery recommendations stay subordinate to canonical identity and taxonomy
• Prefer the smallest structural improvement over new runtime behavior
• Treat titles, tags, browse paths, and related links as distinct signals rather than interchangeable fixes
• Fail closed on changes that would require unsupported redirects, analytics, or external telemetry
• Keep the skill recommendation-only in MVP

Procedure

Step 1: Review current discovery signals

Inspect titles, aliases, tags, browse paths, namespace placement, and nearby index or related-page context. Focus on structural causes of poor retrieval.

Step 2: Identify the weakest discovery link

Classify the primary issue as one or more of:

• title mismatch
• tag quality problem
• browse-path or category weakness
• missing cross-link or related-page path
• index/topology follow-up needed

Step 3: Recommend the smallest structural improvement

Prefer narrow, contract-aligned actions such as retitling for canonical clarity, refining tags, improving browse_path, or requesting a governed index refresh after structural changes are accepted.

Step 4: Mark execution boundaries

If the recommendation needs deterministic execution, route it through sync-knowledgebase-state, review-wiki-plan, or other existing governance surfaces instead of inventing new automation here.

Boundaries

• Do not add search telemetry systems, KPI jobs, or external-service integrations in MVP
• Do not override canonical identity just to chase keyword coverage
• Do not treat tags as a replacement for information architecture
• Do not add a new repo-level script tree for discovery work

Verification

• Recommendations point back to taxonomy and metadata contracts
• Discovery issues are expressed as concrete structural gaps
• Proposed changes stay within MVP and ADR-007 boundaries
• Any execution follow-up is delegated to existing deterministic surfaces
• No unsupported freshness, KPI, or crawler runtime is introduced

References

• schema/taxonomy-contract.md
• schema/metadata-schema-contract.md
• docs/architecture.md
• docs/ideas/wiki-curation-agent-framework.md
• AGENTS.md