haipipe-paper-enter

name	haipipe-paper-enter
description	Open the Paper Console for a paper repo. Use for `/haipipe-paper`, `/haipipe-paper enter <paper-path>`, `/haipipe-paper status [paper-path]`, or when starting work in an existing paper folder. Derives current state from disk (not stored status), renders an open-needs dashboard with the lifecycle frontier, maturity, stable assets, claim/display/round gaps, loopback diagnosis, and next commands, records session state in .paper-console.yaml, and routes free-form follow-up input through the lifecycle in copilot mode.
argument-hint	[paper-path] [free-form input]
allowed-tools	Bash, Read, Grep, Glob, Write, Skill
metadata	{"version":"2.1.0","last_updated":"2026-06-22","summary":"Paper Console: derive-from-disk dashboard + lifecycle router.","changelog":["2.1.0 (2026-06-22): dashboard leads with pitch summary + stage strip before operational details; read order prioritizes 1-pitch.tex; return contract enforces structured tail + failed status; stale-deliverable flag from ref/tex-quality.md.","2.0.0 (2026-06-22): reframed as the Paper Console; added derive-from-disk frontier, free-form routing, copilot policy, and .paper-console.yaml session state.","1.2.0 (2026-06-21): open-needs paper session loader."]}

haipipe-paper-enter (Paper Console)

Open a concrete paper folder as the Paper Console: a context-aware working session for one active paper. It mirrors the Probe Console.

The console:

1. resolves the paper root
2. derives current state from disk, not from stored status
3. renders a dashboard panel (lifecycle frontier + maturity + open needs)
4. records session state in .paper-console.yaml at the paper root
5. routes later free-form user input through the lifecycle

The main job is to expose the paper's current debt board: open claim gaps, display/table gaps, paragraph-placement gaps, round todo gaps, and evidence needs that may require probe/discover/task/insight work. The user often does not know the next stage in advance; the dashboard makes the next need visible.

Follow-up paper actions in the same session must treat that dashboard, especially current_layer, next_layer, and open needs/gates, as the working context. A fresh Claude/Codex session should run enter again.

Story ownership rule: this paper owns its own story, claim wording, narrative, displays, and minimap. Shared evidence lives in project-level probes, discoveries, tasks, and insights. Do not look for or require a project-level narrative layer.

Read first:

../../PHILOSOPHY.md
../../ref/lifecycle-map.md
../../ref/paper-dashboard.md

Then, when the task touches lifecycle shape or rounds:

../../ref/paper-lifecycle.md
../../ref/paper-rounds.md
../../ref/paper-skill-structure.md

When creating or interpreting explicit need records, use:

../../ref/delivery-need.md

Input

Accept either:

<paper-root>

or any path inside a paper root. If no path is supplied, use the current directory.

Resolve Paper Root

Look upward from the supplied path until one of these signatures is found:

• STATUS.md
• 0-lifecycle/
• 0-*.tex and 0-sections/
• 1-compile.sh and 0-sections/

If no paper root is found, report status: blocked and suggest:

/haipipe-paper prospectus "<paper-path>"
/haipipe-paper-lifecycle folder "<paper-path>"

Read Order

Read only files that exist, in this order:

1. STATUS.md
2. 0-lifecycle/README.md 2b. 0-lifecycle/1-pitch/1-pitch.tex — HIGH PRIORITY for dashboard header. Extract the \section*{One-Minute Pitch} paragraph and the \section*{Hook} paragraph. These become the 2-3 sentence "what this paper is about" summary at the top of the dashboard. If the file does not exist or lacks these sections, the dashboard says "pitch not yet written".
3. Stage TeX files (remaining):
   • 0-lifecycle/0-seed/0-seed.tex
   • 0-lifecycle/2-claims/2-claims.tex
   • 0-lifecycle/3-narrative/3-narrative.tex
   • 0-lifecycle/4-display/4-display.tex
   • 0-lifecycle/5-minimap/5-minimap.tex
4. Explicit need records in lifecycle TeX comments or markdown tables. Search for NEED, GAP, TODO, blocked, missing, and open.
5. 0-displays/README.md
6. 0-displays/*/README.md
7. 0-sections/README.md
8. 0-sections/*.tex names and short headers/comments only; do not read full long sections unless needed to diagnose minimap drift.
9. 1-rounds/latest.md, then the referenced round README, discussion.md, decisions.md, todo.md, and applied.md if they exist.
10. Git state:

• git status --short --branch
• git log --oneline --max-count=3

Diagnosis Rules

Derive the current layer from disk, following ../../ref/paper-dashboard.md. Read STATUS.md only as a hint: a stage is done only when its .tex resolves on disk with real content (not the scaffold stub). The frontier is the first stage whose disk predicate fails. If STATUS.md claims more progress than disk shows, flag DRIFT and trust disk.

Per-stage inference when disk is the source of truth:

Evidence	Current layer
only README.md / prospectus lifecycle	0-seed or 1-pitch
1-pitch.tex exists but claims are absent/thin	1-pitch -> 2-claims
claims exist but narrative is absent/thin	2-claims -> 3-narrative
narrative exists but display units are missing	3-narrative -> 4-display
display plan exists but display units/canonical PDFs are missing	4-display
display units exist but paragraph placement is missing	5-minimap
minimap exists and displays are placed	ready for section edit/build/review

Infer maturity separately from current layer:

Evidence	Maturity
seed/pitch only	prospectus
lifecycle + sections + compile script	scaffold
explicit claim ledger	claim-ledger
display map exists	display-map
minimap maps paragraph jobs	section-map
section prose compiles	draft
checks/audits mostly pass	submission-candidate
active round after external/coauthor review	revision

Need diagnosis is separate from lifecycle layer. Extract open needs from:

Surface	Typical need
2-claims GAP/weak/unsupported rows	probe, discovery, task, insight
4-display missing display units	display or task
5-minimap empty paragraph/display slots	paper minimap or display
section comments/TODOs	paper edit or evidence need
round todo.md unresolved items	paper edit, probe, display, citation

Classify each open item using the delivery-need interface:

probe | discovery | task | display | insight | paper-edit

Loopback diagnosis follows the paper lifecycle:

Symptom	Return to
wording, citation, format, stale number	edit cycle
paragraph has no job	5-minimap
figure/table unclear or lacks source/caption/preview	4-display
unsupported or too-strong claim	2-claims / 3-narrative
story not compelling or abstract/intro disagree	1-pitch
paper no longer viable	0-seed

Output Format

The dashboard leads with WHAT THE PAPER IS ABOUT, then WHERE IT STANDS, then WHAT TO DO NEXT. Operational details come after orientation.

Render the stage strip deterministically with the helper, never hand-typed:

sh "$CLAUDE_SKILL_DIR/../../ref/stage-strip.sh" <paper-root>

It prints one line driven by STATUS.md current_layer, e.g. seed ✅ pitch ✅ … → write/edit 🔥 → review ⬜. This strip appears twice: once near the top (orientation) and once as the VERY LAST LINE of the reply (closing every reply in the session, not just the first dashboard; see the orchestrator's "Stage Strip" rule).

The enter skill reads ref/tex-quality.md and flags any stage whose .tex is newer than its .pdf as a stale deliverable in the Open Needs section.

Body order — sections MUST appear in this sequence:

## 📄 Paper Identity

| Field | Value |
|---|---|
| Paper | <title from STATUS.md> |
| Venue | <venue from STATUS.md> |
| Path | ... |
| Branch | ... |

## 📖 What This Paper Is About

<2-3 sentence summary distilled from the \section*{One-Minute Pitch} paragraph
and the \section*{Hook} paragraph of 0-lifecycle/1-pitch/1-pitch.tex.
If no pitch exists, print: "Pitch not yet written — run /haipipe-paper pitch.">

## 📊 Stage Progress

seed ✅  pitch ✅  claims ✅  narrative 🔥  display ⬜  minimap ⬜  →  write/edit ⬜  →  review ⬜

## 🔎 Current State

| Field | Value |
|---|---|
| Current layer | ... |
| Next layer | ... |
| Maturity | ... |
| Active round | <vYYMMDD or none> |

## ✅ Stable

- ...

## 🧾 Open Needs

| Need | Type | Source | Suggested route |
|---|---|---|---|
| ... | probe/display/discovery/task/insight/paper-edit | ... | ... |

## 🔁 Loopback Diagnosis

- ... (omit if none)

## 🎯 Recommended Next

1. `/haipipe-paper-lifecycle ...`
2. ...

## 📦 Artifacts Read

- ...

(return-contract tail here)

seed ✅  pitch ✅  claims ✅  narrative 🔥  display ⬜  minimap ⬜  →  write/edit ⬜  →  review ⬜

The stage strip is the VERY LAST line, placed after the return-contract tail. Keep the dashboard concise. The goal is to orient the session, not to rewrite the paper.

Free-form Routing

After the dashboard, route follow-up input through the lifecycle using the command map in ../../ref/lifecycle-map.md:

seed                       -> /haipipe-paper seed       (haipipe-paper-seed)
pitch / story / sell       -> /haipipe-paper pitch      (haipipe-paper-pitch)
claim / claims / ledger    -> /haipipe-paper claims     (haipipe-paper-claims)
narrative / arc            -> /haipipe-paper narrative  (haipipe-paper-narrative)
figure / table / display   -> /haipipe-paper figures    (haipipe-paper-display)
minimap / paragraph plan   -> /haipipe-paper minimap    (haipipe-paper-minimap)
write / draft / edit / polish -> write/edit skills
review / audit / gate      -> review skills
round / todo / decisions   -> round skills
rebuttal / respond         -> haipipe-paper-rebuttal
slides / poster            -> present skills

If the input does not name a stage, route to the current frontier from the dashboard. If the input is ambiguous, ask before acting.

Copilot Policy

Default mode is copilot. The console may automatically read files, summarize the frontier, classify input, draft or revise a stage .tex, plan section work, and suggest routes.

It must ask before:

calling costly task/PHI/full-data work
committing a claim verdict or downgrading a claim
editing prose across many sections at once
compiling-to-submit or packaging a submission
opening or closing a revision round destructively
filing insight memory as accepted knowledge

Session State

Record the console session at the paper/project root (the nearest directory containing the paper folder), not necessarily the repo root:

.paper-console.yaml

Suggested fields:

paper_root: <path>
active_paper: <Paper-Name>
current_layer: <frontier stage>
maturity: <maturity rung>
active_round: <vYYMMDD or none>
open_needs: <count>
updated: <YYMMDD>

This is session state, not manuscript content. A fresh session re-derives it from disk.

Return Contract

Every reply from a paper specialist (and every enter dashboard) MUST end with the structured tail block followed by the stage strip as the very last line. This is enforced by the orchestrator; omitting it is a protocol violation.

status:        ok | blocked | failed
paper_root:    <path>
current_layer: <layer>
next:          <suggested command>

seed ✅  pitch ✅  claims ✅  narrative 🔥  display ⬜  minimap ⬜  →  write/edit ⬜  →  review ⬜

The status field uses three values: ok (dashboard rendered, session ready), blocked (missing paper root or unresolvable state), failed (read error or inconsistent disk state). The stage strip on the final line is rendered by the helper script, never hand-typed.