Source the task readiness parsers from the projection tree

Status: closed/done · Impact: medium · Complexity: medium

Goal

The readiness consumers each carried their own fence-aware H2 walk and manual table splitting — duplicated markdown parsing alongside the contract. Source them from the always-available parse(text) projection tree (which is ungated even when the typed doc isn’t), deleting the duplicate walks while keeping behavior identical.

Acceptance criteria

• parse-touchpoints, gap-report, and resolve-touchpoints source their sections and table rows from the parse(text) projection tree.
• The three duplicate fence-aware H2 walks and the manual table-splitting are deleted.
• Behavior is identical — no test changes needed and the gate suites stay green.
• The task-specific layers (Location grammar, four-way kind classification, live-codebase resolution, pinned errors) are preserved on top of the tree.

Out of scope

• Migrating scan-placeholders (the last duplicate walk) — deferred until the library ships typed cells with inline-span positions.

---

← Back to Tasks