Stage 4 of the intervention lifecycle (venue-GATED: required for dashboard/ui-card/report, optional for email, skipped for sms/push/reminder/checklist — per STATUS.md stages_skipped). Answers 'what content element carries each claim, and what job does each unit do?' Maps claims to display units (panels, widgets, charts, sections) with per-unit jobs and evidence anchors — the retired minimap stage's concern lives here. Output: 0-lifecycle/4-display/4-display.md + _LOG + 1-probes/. Trigger: display, content elements, panels, widgets, unit jobs, /haipipe-application display.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
A direct command skips the review prompt. Inspect the source before running it.
Stage 4 of the intervention lifecycle (venue-GATED: required for dashboard/ui-card/report, optional for email, skipped for sms/push/reminder/checklist — per STATUS.md stages_skipped). Answers 'what content element carries each claim, and what job does each unit do?' Maps claims to display units (panels, widgets, charts, sections) with per-unit jobs and evidence anchors — the retired minimap stage's concern lives here. Output: 0-lifecycle/4-display/4-display.md + _LOG + 1-probes/. Trigger: display, content elements, panels, widgets, unit jobs, /haipipe-application display.
argument-hint
[intervention-path]
allowed-tools
Bash, Read, Write, Edit, Grep, Glob, Skill
metadata
{"version":"0.4.8","last_updated":"2026-07-19","summary":"Display stage (stage 4, venue-GATED + venue-ALIGNED) — maps each claim to a display UNIT with a required per-unit Job (minimap absorbed); materialization raised as a section in the flat probe pool 1-probes/, uniquely commissioned by this stage. History: ./CHANGELOG.md."}
Skill: haipipe-application-display
Stage 4 of the intervention lifecycle — venue-ALIGNED and venue-GATED.
It answers: what content element carries each claim, and what job does each unit do for the reader?
2-pitch what this intervention sells
3-narrative how claims compose into the output's arc
4-display what element carries each claim + its job <- THIS STAGE
5-section-edit does each section's prose do its job
Read first: ../../../PHILOSOPHY.md, the probe layer's ../../../2-phase/1-probe/haipipe-application-probe/ref/per-stage-dispatch.md (the 4-display lane wording).
What's special: three things make a display a display
1. Claims become UNITS, each carrying a required Job — the minimap absorption.
Every display unit carries FOUR fields: Type, Claim, Job, Data source.
The Job is one sentence on what this unit must make the reader see or do — the retired minimap stage's job-assignment concern, folded in per-unit.
Same role as paper's display; here a unit is a panel, widget, chart, or section rather than a figure or table.
2. Venue-GATED and venue-ALIGNED.
Read STATUS.md | stages_skipped |: display is skipped for sms/push/reminder/checklist (the venue template already fixes the elements), required for dashboard/ui-card/report, and optional venues (email) pull it in on user request.
The available element types come from the venue stage doc's Artifact Principles (0-lifecycle/2-venue/2-venue.md); a venue change re-runs the display set.
3. Display is the ONE stage that commissions its own units.
A missing unit raised from narrative or a section is NOT commissioned there — it becomes a request row this stage later fulfills, and that section closes answered-local.
Only THIS stage commissions render/materialization work for its own accepted units; every other stage's display lane merely LINKs what landed.
A unit whose data source does not exist yet is an evidence need: it is raised as a question SECTION in the flat probe pool 1-probes/PPNN_<topic>/ and dispatched through PROBE — display never runs /haipipe-task or renders inline (LAW 1).
The rendered output lands task-side; the unit's Data source field points at it.
The four phases, in display
DRAFT read 3-narrative.md (the arc, if it fired), 1d-advice.md (what each element carries), 1c-claims.md
(the evidence anchor behind each element), and the venue profile's element types; map every
primary claim to >=1 display UNIT, each with Type / Claim / Job / Content / Data source
PROBE one worker call; an unmaterialized data source is raised as a SECTION in 1-probes/ and, uniquely,
commissioned by this stage; the display lane LINKs landed units. Routing is the probe layer's:
../../../2-phase/1-probe/haipipe-application-probe/SKILL.md (4-display lane in ref/per-stage-dispatch.md)
REVISE unit-set coherence: one job per unit, no orphan units, types match the venue's element set
CHECK the done list below vs the unit set -> Gate Ledger row in STATUS.md
Display PLANS and LINKs units; it never computes, renders, or hand-authors an asset (LAW 1) — materialization is task work reached through the PROBE phase, which binds each question to a QA file in the task/discovery bank.
The artifact
0-lifecycle/4-display/4-display.md — full skeleton in ref/display-template.md (the per-venue dashboard/report unit templates live there, not inline):
Display units one U<nn>: Type + Claim (via A<n> where an advice entry drives it) + Job + Content
+ Data source · Status: planned | commissioned (PP<nn>) | landed
Unit -> section (sectioned venues) which unit goes in which section, and why
mapping
Q-consumer materialization questions, one `## Q-Disp-<n>` block each (Ask / Why / Answer); the display lane LINKs what landed
Sidecar: _LOG_4-display.md (phase journal).
Evidence questions live in the flat probe pool 1-probes/PPNN_<topic>/ (one file per TOPIC; each ENTRY is one ## QX<n> q-executor carrying ### q-executor / ### q-consumer / ### bank binding / ### a-executor), states planned | commissioned | answered | read | answered-local | failed; the stake stays in this doc's Q-consumer.
Formatting: ===== title / ----- sections; content uses no #, Q-consumer questions use ## Q-Disp-<n>; one sentence per line; the doc reads 2-venue.md's Artifact Principles for the available element types.
Definition of done (read at CHECK)
[ ] 4-display.md exists (when the venue requires it)
[ ] every primary claim has at least one display unit
[ ] every unit has all four fields — Type, Claim, Job, Data source
[ ] unmaterialized data sources have 1-probes/ sections (task-routed, commissioned by this stage)
[ ] display types match the venue's available element types
Questions this stage typically raises
DRAFT's RAISE+PLAN step raises what the draft cannot answer. These are the kinds this stage is prone to — read this list, then walk the draft against it.
📤 evidence exists Does a task-produced result exist that this element would show?
An element with no producing task cannot be refreshed.
🎨 element form Panel, chart, table, or plain sentence — which does this claim
need? Ask when the venue's budget makes it a real trade.
📐 venue budget How many elements does this modality allow, and what is cut?
🖥️ render context On what screen / in what client will this be seen? An element
that breaks on a phone is not done.
Exits
promote -> /haipipe-application section-edit (sectioned venues)
or -> /haipipe-application draft (compose the artifact)
End every reply with the closing block (stage line via ../../../haipipe-application/stage-strip.sh).
