Filed by the triage seat (session session_01JEhkB6ShK2HHDzozy7Qwgz, 2026-08-19) under the recorded exit in the pm-dispatch skill's escalation section: "收件箱持续 >20 张 open,或刷新节奏到达每日 ⇒ 把提取冻结进 scripts/pm/ rollup 脚本" — the trigger is met, so this card records it rather than redesigning anything in-session.
Measured trigger (2026-08-19 ~06:1xZ)
- objectstack
needs-user-decision open: 8 - objectui: 29
- objectos: 5
- Total: 42 — persistently above the 20-card line (it was ~33 two rounds ago and has only grown).
The digest (#9822) is currently refreshed by hand each triage round: label queries per repo + literal grep for the os-decision-facets marker + verbatim transcription into the pinned body. At 42 cards the manual transcription no longer fits a round's budget — 23 objectui legacy cards have been queued for transcription across four consecutive rounds without being absorbed.
Scope
A scripts/pm/ rollup script that: queries needs-user-decision across the in-scope repos; extracts each card's facet block by the literal os-decision-facets marker text (grep for the escaped spelling too — sanitizer variants exist, see the platform-readings notes); emits the digest body (blocks verbatim, zero paraphrase; cards without a block listed in an "awaiting backfill" section); the triage seat then pastes/PUTs the generated body into #9822. One-board rule holds: the script generates a view, authority stays on the cards.
Constraints from standing rulings: blocks are transcribed verbatim (2026-08-18 「同意」— standardized block + batch channel); the digest is a generated view, never a second tracker; the batch-adjudication instructions in the header must spell pairs as #9807 B (no angle-bracket placeholders — sanitizer eats them).
domain:skills (SUBJECT is the PM decision-digest protocol; the script is its mechanical half). Unassigned — dispatch via the skills lane.
Filed by the triage seat (session
session_01JEhkB6ShK2HHDzozy7Qwgz, 2026-08-19) under the recorded exit in the pm-dispatch skill's escalation section: "收件箱持续 >20 张 open,或刷新节奏到达每日 ⇒ 把提取冻结进scripts/pm/rollup 脚本" — the trigger is met, so this card records it rather than redesigning anything in-session.Measured trigger (2026-08-19 ~06:1xZ)
needs-user-decisionopen: 8The digest (#9822) is currently refreshed by hand each triage round: label queries per repo + literal grep for the
os-decision-facetsmarker + verbatim transcription into the pinned body. At 42 cards the manual transcription no longer fits a round's budget — 23 objectui legacy cards have been queued for transcription across four consecutive rounds without being absorbed.Scope
A
scripts/pm/rollup script that: queriesneeds-user-decisionacross the in-scope repos; extracts each card's facet block by the literalos-decision-facetsmarker text (grep for the escaped spelling too — sanitizer variants exist, see the platform-readings notes); emits the digest body (blocks verbatim, zero paraphrase; cards without a block listed in an "awaiting backfill" section); the triage seat then pastes/PUTs the generated body into #9822. One-board rule holds: the script generates a view, authority stays on the cards.Constraints from standing rulings: blocks are transcribed verbatim (2026-08-18 「同意」— standardized block + batch channel); the digest is a generated view, never a second tracker; the batch-adjudication instructions in the header must spell pairs as
#9807 B(no angle-bracket placeholders — sanitizer eats them).domain:skills(SUBJECT is the PM decision-digest protocol; the script is its mechanical half). Unassigned — dispatch via the skills lane.