name: plan-md-rows-stay-thin description: "Don't accumulate per-sub-phase detail in PLAN.md phase-table rows; put it in PLAN-P-progress.md" metadata: node_type: memory type: feedback originSessionId: 8e50f1fc-637a-4fa0-bb0f-938d2dc37b09#

PLAN.md's phase table rows must stay thin: a phase row holds only the module list, a <br> pointer to the per-cluster progress file, a high-level "so far / still C" status, plus the approach + test-gate columns. Per-sub-phase ("P6a DONE: …") detail does NOT belong in the row.

Each phase has a split-out summary file: PLAN-P5-progress.md, PLAN-P6-progress.md, … (one-line-with-gates bullet per cluster), and docs/progress-log/p<N>.md (full multi-bullet entries). When finishing a cluster, append the one-liner there — never grow the PLAN.md row.

Why: rows had ballooned to ~8 KB on a single line (P5, then P6) before being split; the user explicitly wants them kept thin from the start.

How to apply: when porting a new cluster, write its summary bullet to PLAN-P<N>-progress.md and the detailed entry to docs/progress-log/p<N>.md. Only touch the PLAN.md row to update the high-level status line, not to add a new "Px DONE: …" blurb. See [[commit-one-test-per-commit]] and the command-cluster-port skill's step 8.