Changelog

3.4.0 — Expand the musical scene horizon

• Add research-backed leads and documented guardrails for pre-1975 Lao and Saigon music, Hindi rock, qawwali, Indigenous North American/Australian guitar scenes, Memphis cassette rap, distinct internet sigilkore, Indonesian punk/hardcore, Khmer and other non-English psychedelia, and nonbinary country.
• Keep distinct timelines, Nations, traditions and artist identities separate; the listener's Lao "Royalist/Ban-Na" terminology is not presented as a verified historical genre.
• Broaden open-ended go discovery without quotas or sonic stereotypes; add source links and manual scene-specific acceptance prompts.
• No change to the DeBarge lottery, Spotify-first setup, platform publishing logic, or track-level verification requirements. These are candidate leads, not a live listening win.

3.3.5 — Spotify-first, zero-interview setup

• Default listening and publishing preferences to Spotify without a required service-selection question; keep setup change ... for explicit overrides.
• Check live writer availability separately from provider preference; never silently publish to a different service just because Spotify cannot write on the current host.
• Document that Gemini's regular Spotify connector supports search/play, not playlist creation; offer an explicit YouTube Music change for Gemini playlist generation and clarify separate library saving.
• Add listener-reported successful Meta Muse + Spotify flow to the manual platform checks without claiming universal account access or exact track/order fidelity.
• Preserve four-axis taxonomy, playlist voice, length defaults, DeBarge behavior and streaming authorization safeguards.

3.3.4 — Distinguish generated previews from saved playlists

• Correct Claude field evidence: two playlist results were displayed but did not save to Spotify; report 0 confirmed saved rather than partial publication.
• Keep generated previews, durable saved playlist objects, and exact item/order verification separate. Premium is required for personalized prompt-generated playlists via Claude, but does not guarantee exact duplicate insertion.
• Preserve the DeBarge selection and all prior music curation behavior.

3.3.3 — Reconcile creations before reporting a total failure

• Initial field report described two displayed playlist results without descriptions; later user clarification established neither was saved. See 3.3.4 for corrected saved-state guidance.
• Introduce stage-by-stage publication reconciliation, separating playlist creation, contents, metadata, provider denials, and uncertain results. Never create a third playlist before resolving previous attempts.
• Preserve DeBarge odds, original 25-entry sequence and one-roll-per-new-playlist behavior.

3.3.2 — Spotify refusal is not an invitation to invent tools

• Differentiate Premium-gated prompt-generated Spotify playlists from exact ordered-track editing in Claude.
• Preserve 25-entry DeBarge selections and report 0 saved when the official Spotify connector refuses publication; never reroll or force mobile manual duplication.
• Avoid making claims about actual account subscription or suggesting unavailable add_playlist_items tools; add a field-report regression matrix.
• DeBarge probability and special result unchanged.

3.3.1 — Make the liner notes worth reading

• Distinguish playlist title, proposed streaming description, and track-specific liner notes in the core skill rather than relying solely on a generic "be snarky" persona.
• Add concrete flat-versus-expressive examples, a succinct copywriting format, and manual Claude/ChatGPT quality-parity checks without inventing audio evidence.
• Clarify provider boundaries: a distinctive Suggested playlist description in chat does not mean Claude's Spotify connector saved that metadata.
• Add Claude-specific rewrite-in-place troubleshooting for already-curated playlists; do not re-curate, republish, or reroll DeBarge on a writing-only revision.
• Keep discovery taxonomy, mood modifiers, setup, publishing authorization and the protected DeBarge Protocol unchanged.

3.3.0 — The setup command

• Claude/Muse distribution follow-up: shorten the Agent Skills discovery description to Claude's 200-character guidance, publish Claude Skill and Plugin ZIPs alongside Gemini, and document consumer Muse versus Muse Code with native Spotify integrations and verified permission boundaries.
• Add first-class setup, setup status, setup change ... and setup again to the main skill and existing setup skill, with service discovery, listening/publishing preferences and clear defaults.
• Apply a listener-approved, actually accessible Hipster Music Setup v1 profile to future go requests while rechecking live provider write capability.
• Keep persistence honest: use actual user-authorized host storage if supported, otherwise display a portable session-only profile; never leak personal preferences into public skill packages or claim persistent cross-chat memory.
• Separate Gemini's thematic YouTube Music generation / Save to library from exact recording insertion; never turn setup into a Spotify OAuth bypass.
• Add manual setup regression cases and cross-host instructions; keep curation taxonomy, voice and DeBarge Protocol unchanged.

3.2.4 — Gemini's YouTube Music routing is not ChatGPT's

• Remove the cross-host fallback that told Gemini users to select two plugins in a ChatGPT-style composer.
• Distinguish Gemini's @YouTubeMusic playlist generation, exact curated track insertion, and the Save to library step; disclose approximations rather than inventing exact results.
• Preserve original tracklists for a single Gemini-native recovery step, without re-curation or repeat DeBarge roll.
• Add Gemini-specific regression tests and setup guidance; music taxonomy, snark and DeBarge odds unchanged.

3.2.3 — Make Gemini's record clerk unmistakably insufferable

• Promote the arch, self-important persona to the beginning of canonical skill activation and the generated Gemini Apps SKILL.md, before streaming setup instructions.
• Require a dry opener, several track-specific witty asides, and a smug closing; factual music/provider status still takes precedence.
• Explain how to reload a GitHub-linked session or replace a saved Gemini skill using the whole five-file package; static tests and manual tone cases added.
• Preserve taxonomy, all go modes, DeBarge's original protected protocol and the publishing workflow.

3.2.2 — Go sad, obviously

• Add go sad: musically melancholy, wistful or heartbroken, but not synonymous with low tempo, minor keys, acoustic sound or ambient music.
• Support go sad fast, go sad weird, and go sad chill as natural combinations without changing the four-axis discovery method.
• Add examples, manual regression checks and README quick-start commands.
• Keep go sad eligible for the existing one-in-100 DeBarge Protocol; its trigger, odds and exact 25-track outcome remain unchanged.
• Preserve careful publishing verification and the snarky persona without mocking real grief.

3.2.0 — An extremely important logo, naturally

• Rebuild the reference crimson faceted five-point star as real transparent, square SVG vector artwork with clean negative-space cutouts.
• Add a simplified composer icon and a subtler, vinyl-etched signature logo, and reference both in the portable OpenAI plugin manifest for light and dark modes.
• Add a brand guide and regression tests for file references, square SVG dimensions, transparent backdrop, and correct naming.
• Keep all musical taxonomy, go flags, publishing behavior and DeBarge odds unchanged.

3.1.9 — The Insufferable Record-Store Clerk

• Give Hipster Music Generator an overtly self-important, passive-aggressive and snarky comic voice across introductions, track transitions, feedback and endings.
• Keep track-specific sonic explanations, trustworthy artist/title/version claims and technical publish reports first; on errors give clear status before the customary jab.
• Sarcasm targets the curator's own ego, stale algorithms and playlist clichés—not real musical cultures or vulnerable listeners. Honor less snark.
• Extend style examples, ChatGPT/Gemini instructions and tone regression checks without changing sonic taxonomy, routing, go modifiers or the DeBarge Protocol.
• No claim that textual CI can measure whether the joke lands; user listening feedback remains essential.

3.1.8 — DeBarge across every go modifier

• Fixed an eligibility conflict introduced with go short, go long and numeric track-count modifiers: ordinary length requests no longer accidentally disable the one-in-100 Easter egg.
• On a genuine single-roll success, 25 repeated original recordings override all normal mood, speed, short/long and numeric count goals.
• Explicit no-DeBarge, unique-song, or unambiguously non-overridable length constraints still prevent a surprise; no second roll on publishing retries.
• Added manual acceptance matrix and contract tests. Randomness remains contingent on an actual reliable random tool, and streaming provider duplicate support remains unverified.

3.1.7 — Expressive go modifiers and length controls

• Added orthogonal mood/speed instructions: go fast, go angry, go chill, go slow, go weird, plus combinations and natural-language mood requests.
• Keep the full four-axis taxonomy; modifiers change sonic properties, not force genre quotas or invented BPM data.
• Preserve default 12 tracks and introduce explicit go short (6), go long (24) and numeric requested lengths, with 50 as a soft quality target rather than hard/schema limit.
• Require clear reporting when providers or hosts produce fewer items than requested; no silent truncation.
• Added manual cross-host mood/length acceptance cases and structural tests; live fidelity and maximum host throughput remain unverified.
• DeBarge odds, trigger and 25-entry outcome unchanged.

3.1.6 — Publishing-first routing experiment

• Move the connected-service publishing preflight ahead of the seven-step music-curation workflow for go; do not create an empty test playlist.
• Make Spotify/YouTube Music playlist creation prominent in skill activation metadata and plugin discovery text.
• Curate service-playable real recordings after checking accessible write capabilities; retain complete taxonomy and verification requirements.
• Keep the two-plugin Spotify success case as a manual A/B control. No claim that host-controlled automatic activation is fixed.
• Preserve the DeBarge Protocol unchanged.

3.1.5 — Preserve curated playlist when app routing fails

• Record the observed ChatGPT Android limitation: a go with the curator alone did not expose Spotify's write action, while explicit dual-plugin selection saved a 12-track playlist.
• When writer activation fails, retain the exact ordered playlist and instruct the user how to publish it in a follow-up with the provider selected—without re-curating or rerolling any Easter egg.
• Distinguish a connected account from a writing tool missing in the current conversation. Automatic cross-plugin activation remains unverified and not guaranteed.
• Maintain the DeBarge trigger, odds, and 25-entry special behavior unchanged.

Unreleased — One-command streaming handoff

• Make go request an automatically saved playlist through an accessible, authorized music writer, when one is available.
• Choose the sole writable provider automatically, prefer a grounded service preference, and ask once only if multiple choices remain.
• Preserve host routing limitations, account permissions, privacy, and approvals; don't promise automatic access to other skills.
• Add go, text only opt-out and clear no-publisher fallback.
• Add manual Spotify/ChatGPT and YouTube Music/Gemini routing tests; actual host orchestration remains unverified.
• DeBarge Protocol unchanged.

Unreleased — Listening-quality iteration (experimental)

• For open-ended go, select among two or three musically credible concepts rather than jumping to the first regional pairing.
• Require track-specific sonic reasons and honest uncertainty when recordings could not be heard or independently described.
• Accept lightweight numbered feedback, using only a playlist actually available in the conversation.
• Request post-publish reconciliation of intended versus provider-confirmed tracks, mixes, and order; unknown remains unknown.
• Added manual ChatGPT/Spotify and Gemini/YouTube Music regression cases. Provider fidelity and musical superiority are not yet established.
• Preserved the DeBarge Protocol's random trigger, exact recording and 25-entry behavior without change; made the public-facing README allusion less explicit.

3.1.0 — Portable plugin packaging and discovery setup

• Added portable plugin.json, installed curation skill, and onboarding skill.
• Added capability-aware checks for accessible streaming app tools and actual playlist-writing authorization.
• Kept publication optional: missing or search-only integrations fall back to text and portable JSON.
• Added Plugin Creator instructions; legacy ChatGPT Project instructions remain a fallback.
• Added offline plugin structure tests. Actual ChatGPT installation and music-account integrations still require live testing.

Unreleased — ChatGPT Project adapter

• Added phone-friendly ChatGPT Project instructions and one-file knowledge bundle.
• Added reproducible bundle generator, freshness tests, and manual in-Project smoke tests.
• Kept GitHub as the canonical skill source; no claim of native ChatGPT installation or streaming publication.

3.0.0 — 2026-10-16

• Consolidated original and v2 music taxonomy into an Agent Skills-style repository.
• Kept expanded international scene references and sonic-first discovery distance.
• Preserved late-Motown-adjacent camp and the true 1-in-100 DeBarge Protocol.
• Scoped prior-playlist repetition limits to a practical preference instead of a session-wide hard ban.
• Added standardized playlist JSON contract, offline validation and tests.
• Added transparent, not-yet-implemented streaming adapter guides for YouTube, Spotify and Apple Music.
• Added cross-agent installation guide and public repository workflow.