Plan Reads

Plan Reads

Group physical reads without losing point traceability.

Follow ../../references/interaction-contract.md.

Process

1. Require a canonical map with no blocking identity, address, access, width, or final-decoding holds; do not require a separate blanket human approval.
2. Run python3 <skill-dir>/scripts/run.py --input <map.json> --output <read-plan.json> --max-gap <unused-address-count>.
3. Verify block quantity, gaps, interval, target limits, and point slices.
4. Verify automatically that every point appears in exactly one physical block.
5. Verify planning_options and its hash record the gap policy.
6. For final output, verify the plan's canonical-map hash matches the exact map.

Inspect these fields together in the emitted plan. For routine generation, use the wrapper's validation and existing trace records; do not repeatedly browse runtime code or reconstruct validators unless a concrete discrepancy requires it.

Output files

• read-plan.json - Use this machine-readable list of bounded Modbus requests when building a target tool. It also records which points each request returns and why blocks were split.

Completion requires every active point to be planned once with explicit route, unit, area, offset, and width.

Use --max-gap 0 without asking. Ask once only when a nonzero sparse-read policy is needed. Replan automatically after every map change.

Stop

• Stop while blocking identity, address, access, width, or final-decoding holds remain.
• Use --max-gap 0 unless a nonzero sparse-read policy is explicitly required.
• Permit function codes 01 through 04 only.

Handoff

• One target: suggest build-node-red, build-modpoll, or build-modscan based on the named tool.
• Several targets: suggest build-tool-pack.