Tech-lead onboarding

Books / Pearl Prime

bestseller register / PROVEN-AT-BAR (one blind-10 set only)

Exactly one set of ten corporate_managers cells has operator Layer-4 acceptance. The flagship remains system working by agent proxy. Its full-book golden is byte-identical to its snapshot; it was deliberately re-ratified on 2026-08-09 (commit df1d97fc66) to clear an F13 HARD_FAIL, which is a governed update and not drift.

What this pipeline produces

A readable book artifact—book.txt and, when rendered, the customer EPUB—plus its gate reports and named acceptance verdict. A listing is not an EPUB.

Where it really stands

Acceptance layer: bestseller register / PROVEN-AT-BAR (one blind-10 set only)

Evidence: artifacts/qa/LAYER4_10CELL_BLIND_READ_PREP_2026-07-10/LAYER4_VERDICT_2026-08-09.md

Run it yourself

This documented sentinel uses the mandatory four-piece production chord. It requires the repository’s normal dependencies and governed inputs.

python3 scripts/run_pipeline.py \
  --topic overthinking \
  --persona gen_z_professionals \
  --series overthinking \
  --angle analysis_paralysis \
  --arc config/source_of_truth/master_arcs/gen_z_professionals__overthinking__spiral__F006.yaml \
  --teacher ahjan \
  --location nyc_metro \
  --runtime-format standard_book \
  --structural-format F006 \
  --pipeline-mode spine \
  --quality-profile production \
  --exercise-journeys \
  --out artifacts/user_runs/gen_z_nyc_overthinking_teacher_mode/plan.json \
  --render-book

Input: governed arc, teacher, topic, persona, and location. Output: the plan at the cited path plus rendered book artifacts and reports beneath the run directory. A produced book.txt proves only path works until the acceptance layers are evaluated.

The map

Verified path in supplied context	Role
`scripts/run_pipeline.py`	Canonical CLI
`config/source_of_truth/master_arcs/gen_z_professionals__overthinking__spiral__F006.yaml`	Sentinel arc
`phoenix_v4/rendering/book_renderer.py`	Rendered prose
`phoenix_v4/quality/chapter_flow_gate.py`	Layer-1 flow report
`phoenix_v4/quality/book_quality_gate.py`	Layer-1 book quality
`scripts/ci/check_book_story_authored.py`	Research-fit binding signal
`artifacts/qa/snapshots/CANONICAL_FLAGSHIP_BOOK.txt`	Full-book golden — byte-identical; re-ratified 2026-08-09

What is broken right now

Drift detectors is red repo-wide. The named cause is the five-layer exercise gate below; until it is green, no merge is gate-protected.

authoring

check_flagship_exercise_five_layer.py fails on flagship chapters 10, 11 and 12 — not chapter 2. ch10 and ch12 each drop the _If restless or resistant:_ alternative-entry line; ch11 drops two layers including the readiness check.

The traps

Do not re-tune the composer to fix register. Hand-seamed use of the same atoms already read at the target register; the lever is the Layer-3 line-edit lane.
Production builds require --pipeline-mode spine --quality-profile production --exercise-journeys. Do not weaken the chord gate.
A listing is not an EPUB. Catalog metadata must never be counted as a readable customer file.
A gate PASS is at most structurally clear; never collapse it into an unlayered “all checks pass” claim.

Who to ask

Pipeline owner

Pearl_Prime

Coordination

Pearl_PM owns scope, sequencing, and status claims through artifacts/coordination/ACTIVE_WORKSTREAMS.tsv.

Architecture

Pearl_Architect owns architecture decisions and cap rulings through docs/PEARL_ARCHITECT_STATE.md.

Your first week

Run the sentinel with the four-piece chord and preserve book.txt plus every gate report.

Classify its result as path works, structurally clear, authored candidate, or system working—never by green checks alone.

Reproduce the five-layer exercise gate failure on flagship ch10/11/12 and find why those chapters drop layers at render.

Take one failed ONTGP dimension through the line-edit lane and produce a new Layer-3 packet.