Home / Ops / Housekeeping and Process
Updated Jun 21, 2026 · Affirmology_AppChanges_MasterCodePrompt_v1.md
Paste the block below into your Claude Code window on the Mac. It builds all four changes in sequence, each as its own commit with tests, pausing for your review between them. The order matters: C29 schema first (it is the sorting spine), then C28 (uses the schema), then C26 (adds the Structure fields), then C3 (uses C26's bed config).
Guardrails baked in: the locked demo never changes, the QC gate stays, no em dashes, every change is Studio-side and additive, and engine changes are vendored into affirmology-studio so cloud matches local.
Read these first and confirm: Affirmology_PROJECT_STATE.md, Affirmology_BeliefInstallation_CoreCraft_v1.md, Affirmology_BeliefInstallation_C29_Kickoff_v1.md, Affirmology_PractitionersTechniques_ResearchCorpus_Plan_v1.md, Affirmology_AudioPacingActivation_C26_BuildBrief_v1.md, and Affirmology_FrequencySoundLayer_C3_Spec_v1.md.
Build the following FOUR changes in order, each as its own commit with tests, and STOP for my review after each before moving to the next. Hard rules for all four: the locked demo render path must be byte-for-byte unaffected (it never passes --council); keep scripts/audio_qc.py running on every render and the auto-tune ceiling at 1.05; everything is Studio-only and additive with today's behavior as the default; no em dashes anywhere; vendor any engine change into affirmology-studio/affirmology-agent so cloud matches local; run the full studio + engine test suites and show me results; do not run paid ElevenLabs renders inside tests.
STEP 1, C29 the belief-installation tagging spine. Add `rule` (values 1 to 10, multi-valued) and `form` (audio_craft, journaling_prompt, mindset_practice, journey_companion, breath_somatic, frequency_sound) to the craft schema (craft.py, both the engine copy and the vendored copy) and to the craft_seed emission so new entries are tagged. Write a one-time backfill that tags existing craft entries (cheapest model, the existing cost cap). Do NOT run the backfill yet; show me the migration and the backfill script and wait for my go. Add a test that the new fields persist and that search_craft can filter by rule and form.
STEP 2, C28 the practitioners/techniques research lane. Extend the nightly growth cron with a `practitioners` (or `wisdom`) research lane per the plan, seeded from the canon in the plan doc, each entry carrying teacher, lineage, core method, mechanism (evidenced vs traditional vs metaphorical), actionable practice, how it serves an audio, tags, usefulness/confidence, the SCOPE flag (now/future/archive), and the C29 rule + form tags. RESEARCH ORDER: hypnotherapy first (Marisa Peer/RTT as the tier-walled benchmark, Erickson, Elman, Yapko, Gilligan, etc.), then the public-domain teachers that we can ingest FULLY (Neville Goddard, Wattles, Shinn, Troward, Behrend, James Allen, Coue, the Kybalion) before in-copyright authors. Tier-wall copyright (in-copyright and branded programs are benchmark-only, distilled into original method, never reproduced); flag quantum/wave-function framings as metaphor. Keep it under the existing growth-cron caps. Show me the seed wiring and a small dry-run sample before any full run.
STEP 3, C26 audio pacing + activation. Per the brief: add pace (deep/standard/activation), target_minutes, and bed_profile to the Structure model; drive ElevenLabs speed from pace; insert pauses and silence-after-questions (prefer mix-time silence in audio_mix.py if SSML is not reliable); enforce per-type target length so two people on the same structure land within tolerance; wire Hypnos (pace + pause plan + target) and Apollo (pacing gate) in council.py. Add the four acceptance tests from the brief and show me one real deep render and one real activation render to listen to before deploying.
STEP 4, C3 frequency/sound layer. Per the spec: add a synthesized tone-bed stage in audio_mix.py after voice+music and before the master (isochronic + tonal center first, binaural as a headphone-flagged option), driven by Structure.bed_profile with the state-to-frequency table (delta sleep, theta journey, alpha morning, rhythmic activation), low fixed-headroom gain, a Studio-only --tones flag, off by default. Carry the honesty note into the build report. Add the five acceptance tests and show me three real renders (before-bed delta, journey theta, tones-off control) before deploying.
After all four: update Affirmology_PROJECT_STATE.md with what shipped (commits + build tags), and check off C26, C3, C28, and the C29 schema portion in Affirmology_TestAndChange_Queue_v1.md. Give me the diffs, the test output, and the render files to listen to at each stop.
Tip: if Code is slow or you want to ship faster, you can run STEP 3 (C26 pacing) on its own first, since it gives the most audible improvement and C3 builds on its bed_profile field.