playbooks: repo-sweep next loads orchestrate and use-your-skills on every step, even single-detector steps

[kyle-sexton]

Problem

skills/repo-sweep/reference/next.md section 2 step 1 requires every next run to invoke /session-flow:orchestrate and /discipline:use-your-skills before the step runs. Both skills load their full bodies into the session, about 3k tokens together.

A step whose entry runs one deterministic detector does not use either skill. It does no delegation and cannot miss a skill.

Evidence

Sweep of melodic-software/.github (PR #153), step testing-audit (testing:audit@0.9.2):

• The step ran cant-fail-scan.sh once: 1 test file, 25 test blocks, 0 findings, exit 0.
• The step was ticked no findings with no commit.
• Both mandated skills loaded and neither changed anything in the run.

Proposed fix

Make the two invocations conditional in next.md. Skip them when the entry lists a single skill whose run is a script detector. One option is a per-entry catalog flag, for example - prime: false. Keep them for multi-skill entries and entries whose fix path delegates.

playbooks 0.13.0.

[kyle-sexton]

Seen again on the claude-memory step of the melodic-software/.github sweep (PR melodic-software/.github#153): one skill, no subagents, and both preloads (/session-flow:orchestrate, /discipline:use-your-skills, about 5k tokens together) changed nothing in the run.

[kyle-sexton]

Issue stays closed. Audit follow-up in #5295 (Refs only). This comes from the 2026-09-29 audit of the unattended Cursor run.

The audit found the hygiene catalog test did not exercise - prime: false or a non-false value, and next.md did not name the prime column. On the branch, catalog.test.sh covers both, next.md addresses the column as "column 8", and the catalog applies this issue's criterion (single-skill script detector) per entry, with a Notes line for each single-skill entry left primed.