morphiq-rank（形态等级）

Pipeline Position

Step 2 of 4 — consumes morphiq-scan output.

• Input: Scan Report (JSON) from morphiq-scan.
• Output: Prioritized Roadmap (JSON) → consumed by morphiq-build.
• Data contract: See PIPELINE.md §2 for the Prioritized Roadmap schema.

Purpose

Morphiq Rank transforms raw scan findings into an actionable, prioritized roadmap. It determines severity, assigns progressive discovery tiers, calculates priority scores, and controls how many issues are revealed. The output tells morphiq-build what to fix and in what order.

Workflow

Step 1: Ingest Scan Report

Parse the Scan Report JSON. Extract per-page scores, domain-level scores, all identified issues, and the overall pipeline score (0–100).

Step 2: Create Issues

For each finding, create a formal issue:

For fanout issues, severity depends on parent prompt type fan-out depth. site: and citation-producing sub-queries escalate one level.

For fanout-* issues, populate fanout_context from the scan report's query_fanout.simulated_queries[] and query_fanout.suggested_content[]. Each simulated query mapping to this issue becomes a triggering_sub_queries entry:

• query ← from simulated_queries[].query
• model_origin ← from simulated_queries[].model (rename model → model_origin)
• prompt_type ← from simulated_queries[].prompt_type
• citation_weight ← from simulated_queries[].citation_weight
• parent_prompt ← for simulated queries, set to "(simulated)"; for suggested_content[] entries, use the suggestion field

If the Delta Report's content_creation_queue has entries matching this issue, include their competitor_sources in fanout_context.competitor_sources[].

Deduplication: Technical issues hash by brandId + checkCode + pageUrl. AI visibility issues hash by brandId + category + title.

For all issue types and severity logic, read references/issue-catalog.md.

Step 3: Assign Tiers

Edge cases: fanout-wrong-page-type → Tier 2. Multi-tier issues → lowest applicable tier.

For tier definitions and dependency logic, read references/tier-progression.md.

Step 4: Calculate Priority Scores

priority = (severity_weight × 0.4) + (page_impact × 0.3) + (citation_potential × 0.2) + (effort_inverse × 0.1)

severity: critical=100, high=75, medium=50, low=25. page_impact: % pages affected. effort_inverse: low=100, medium=50, high=25.

Step 5: Apply Progressive Reveal

Score-based: <30 → fundamental only, ≥30 → +intermediate, ≥60 → +advanced, ≥80 → all tiers.

Page-based: First run = homepage only. Each subsequent run unlocks one more page (home → pricing → features → product → solutions → about → blog → other → docs).

Backlog cap: Max 10 issues in identified state.

Step 6: Set Dependencies

Cross-tier: T1 → T2 → T3 → T4. Higher-tier issues on same URLs only become actionable when lower-tier issues are resolved. Within-tier explicit dependencies also apply.

Step 7: Produce Prioritized Roadmap

Assemble JSON (PIPELINE.md §2): issues by tier, sorted by priority, with severity, remediation hints, affected URLs, dependencies, and reveal state metadata.

Reconciliation (Re-runs)

On subsequent scans with existing issues:

1. Auto-close issues where the check now passes (unless PR-linked)
2. Escalate worsened issues
3. Create new issues for new findings
4. Detect regressions (previously completed issues re-appearing)

For issue lifecycle and auto-close logic, read references/tier-progression.md.

Reference Files