Design capability-specific code graph scorecards

[forhappy]

Summary

• add Plan 025 for capability-specific code-graph quality scorecards
• separate source fidelity, structural usefulness, and agent utility instead of publishing one global score
• define seven initial capability slices, strict status semantics, independent evidence requirements, and a six-phase rollout
• register the plan and its dependencies in the advisor index

Key decisions

• preserve existing graph, audit, topology, React, and query qualification contracts
• introduce only additive qualification contracts at v1; do not bump product, crate, producer, or framework-pack versions
• keep Graphify and OpenClaw out of the runtime and qualification boundary
• begin real-repository qualification with pinned FastAPI, FastAPI template, and React corpora
• gate expected connectivity and paths from evidence rather than rewarding arbitrary graph density or community counts

Validation

• git diff --check origin/main...HEAD
• reviewed the branch diff to confirm it contains only advisor-plans/025-capability-specific-code-graph-scorecards.md and advisor-plans/README.md

Documentation-only change; no product tests were required.