CHANGELOG.md

Changelog

All notable changes to AI Knowledge Engine will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

instructions_version field in deployed kb.config.yml should match this VERSION. On mismatch, kb_upgrade.py (Phase 4) helps migrate.

🤖 Note for AI agents reading this file: the entries below are historical and may reference scripts/files that no longer exist by those names (e.g., install.sh was renamed to finalize.sh in 0.9.1). For the current canonical state, always rely on what is actually present in knowledge-base/shell/ and the live instruction modules, not on CHANGELOG history. The ## [Unreleased] section reflects pending work only; the latest released version is the source of truth.

[Unreleased]

Added

• Out-of-the-box media processing (transcription, OCR, archives).
  • knowledge-base/scripts/kb_stt.py — speech-to-text for audio/video. Default backend is faster-whisper, which decodes audio via bundled PyAV, so no system ffmpeg is required on macOS/Windows/Linux. Fixes the common macOS Homebrew PATH failure ("ffmpeg not found"). Falls back to openai-whisper when configured. Graceful degradation: missing backend → review queue with an OS-specific install hint.
  • knowledge-base/scripts/kb_ocr.py — OCR for images/scans. Default backend rapidocr-onnxruntime (no system deps); Tesseract optional.
  • Archive unpacking (.zip, .tar, .tar.gz) into raw/unsorted/ for re-ingestion, with a zip-bomb safety cap.
  • knowledge-base/templates/requirements-media.txt — optional media deps.
  • media: section in kb.config.yml (stt / ocr / archives).
  • knowledge-base/15_MEDIA_PROCESSING.md — contract module.
  • kb_doctor.py now reports STT/OCR/ffmpeg readiness.
• knowledge-base/scripts/kb_reindex.py — cross-platform (pure-Python) reindex orchestrator so the pipeline runs on Windows without Git Bash/WSL. The watcher now uses it; reindex.bat delegates to it.
• kb_common.find_ffmpeg() (cross-platform, probes /opt/homebrew/bin etc.) and os_install_hint().
• .cursor/rules/ — repo rules for AI agents (architecture/privacy/ cross-platform + media handling).
• docs/ROADMAP.md — phased roadmap with task checklists
• VERSION file (semver of instructions)
• CHANGELOG.md
• docs/MAINTENANCE.md — contributor rules for keeping instructions/scripts/translations in sync
• Role templates:
  • knowledge-base/examples/psychologist-gestalt.yml
  • knowledge-base/examples/music-video-director.yml
  • knowledge-base/examples/battle-rap-producer.yml
  • knowledge-base/examples/viral-short-form-veo.yml
  • knowledge-base/examples/russian-software-engineering-student.yml
  • knowledge-base/examples/startup-opportunity-explorer.yml

Changed

• Agent instructions now route chat-uploaded files into raw/*/unsorted/ only after user confirmation and require a user question before low-signal files are extracted into knowledge/.
• kb_ingest.py now mechanically handles the stt, ocr, and archive strategies (previously declared but unimplemented), and recognizes many more audio/video/image/archive extensions. Silent except: pass blocks in NLP enrichment replaced with debug logging (--verbose).
• 03_PIPELINE.md and 01_PREREQUISITES.md updated: ffmpeg/tesseract are now genuinely optional (only for alternative backends); the defaults need no system tools.
• CI remains intentionally disabled: this repo is a download-only template that is never deployed, so .github/workflows/ci.yml.disabled stays off and is documented as such (run pytest locally instead).
• scripts/kb_upgrade.py syncs the new scripts (kb_stt, kb_ocr, kb_reindex) and the previously-missing kb_populate.

Translation impact

• README.md role-template table updated; i18n/ru/README.md updated alongside with the same new templates.
• i18n/ru/knowledge-base/06_AGENTS_TEMPLATE.md updated for the new chat-attached file rule.
• i18n/ru/knowledge-base/03_PIPELINE.md updated for the chat attachment staging rule.

[0.1.0] - 2026-05-16

Added

• Initial public structure: knowledge-base/ (13 instruction modules), quick-start/INIT_GUIDE.md, examples/ (3 role templates)
• Bilingual README.md / README.ru.md
• MIT License

[0.2.0] - 2026-05-16

Phase 1 (reference implementations) complete.

Added

• knowledge-base/templates/ directory with all configuration templates:
  • kb.config.yml.template (with full mode_profiles for default/super)
  • repomix.config.json.template
  • AGENTS.md.template
  • KNOWLEDGE_STRUCTURE.md.template
  • DATA_PLACEMENT_EXAMPLES.md.template
  • requirements.txt, requirements-dev.txt
  • .gitignore.template
• knowledge-base/scripts/ directory with reference Python implementations:
  • kb_common.py — shared utilities (frontmatter, hashing, logging, wikilinks)
  • kb_doctor.py — post-deploy smoke test (env, deps, structure, spaCy)
  • kb_ingest.py — full pipeline (raw → processed → metadata, NLP, routing)
  • kb_lint.py — Level 1 health check (11 rules, lifecycle-aware, auto-fix)
  • kb_watch.py — watchdog-based file watcher with polling fallback
  • kb_reflect.py — reflection trigger logic (importance threshold + weekly)
  • kb_nlp_batch.py — incremental NLP re-enrichment
• knowledge-base/scripts/tests/ — pytest suite (59 tests) covering common, lint, ingest, and reflect logic
• knowledge-base/shell/ directory with POSIX-safe wrappers:
  • reindex.sh (with daily consolidation block)
  • watcher.sh (start/stop/status/daemon)
  • lint.sh, doctor.sh
• knowledge-base/00_OVERVIEW.md — agent's deployment map: what to read, what to copy, in what order
• docs/ARCHITECTURE.md — contributor-oriented architecture overview

Changed

• Instruction modules now point to reference implementations: 01_PREREQUISITES.md, 02_INIT.md, 03_PIPELINE.md, 05_INDEX.md, 06_AGENTS_TEMPLATE.md, 09_LINT.md, 13_AUTORUN.md
• knowledge-base/README.md — updated reading order, mentions templates/, scripts/, shell/
• 09_LINT.md — --only flag added to the kb_lint contract

Translation impact

• All instruction modules touched in this release will need re-translation in Phase 2. Russian counterparts in i18n/ru/ (to be created) should be marked as stale until updated.

[0.3.0] - 2026-05-16

Phase 3 (Initial Population Helper) complete.

Added

• knowledge-base/14_INITIAL_POPULATION.md — new module describing role-aware generation of DATA_PLACEMENT_EXAMPLES.md
• placement_examples: section in all three role templates:
  • examples/programmer-senior.yml (7 artifacts + quickstart + don't-drop)
  • examples/marketing-director.yml (8 artifacts)
  • examples/creative-hybrid.yml (8 artifacts)

Changed

• 02_INIT.md — Phase 3 hook directing the agent to 14_INITIAL_POPULATION.md after structure creation
• templates/DATA_PLACEMENT_EXAMPLES.md.template — clarified its role as a starting skeleton replaced during deployment
• knowledge-base/README.md, 00_OVERVIEW.md — reading order includes module 14

Translation impact

• 02_INIT.md, 14_INITIAL_POPULATION.md, README, and 00_OVERVIEW need new Russian translations once Phase 2 begins.

[0.4.0] - 2026-05-16

Phase 2 (localization infrastructure) partially complete: framework + critical entry-point translations. Modules 03–13 and INIT_GUIDE.md remain in Russian (with a "pending English translation" banner) for incremental translation.

Added

• i18n/ru/ directory with all 17 Russian instruction files copied from the canonical sources, each carrying frontmatter: translation_of, source_commit, source_version, translated_at, translator
• scripts/check_translations.py — drift report generator
• i18n/TRANSLATION_STATUS.md — auto-generated drift report (currently all in-sync, will mark stale as canonical EN files evolve)
• docs/TRANSLATING.md — contributor guide for adding/maintaining translations

Changed

• Translated to English (canonical):
  • knowledge-base/README.md
  • knowledge-base/01_PREREQUISITES.md
  • knowledge-base/02_INIT.md
  • knowledge-base/14_INITIAL_POPULATION.md
• Modules 03–13 and quick-start/INIT_GUIDE.md now display a "Pending English translation" banner at the top, with a pointer to the Russian copy in i18n/ru/. Technical contracts inside (file names, code, command names) are language-agnostic and remain usable.
• README.md link now points to i18n/ru/README.md
• README.ru.md is now a redirect page to i18n/ru/README.md

Translation impact

• Once the canonical EN versions of modules 03–13 are written, the existing i18n/ru/ copies will be automatically marked as stale by check_translations.py (their source_commit will fall behind HEAD). This is the intended workflow.

[0.5.0] - 2026-05-16

Phase 2 fully complete: every canonical instruction module is now in English.

Changed

• Translated to English (replacing Russian originals):
  • knowledge-base/03_PIPELINE.md
  • knowledge-base/04_REVIEW.md
  • knowledge-base/05_INDEX.md
  • knowledge-base/06_AGENTS_TEMPLATE.md
  • knowledge-base/07_INTERACTION_LOOP.md
  • knowledge-base/08_PORTABLE.md
  • knowledge-base/09_LINT.md
  • knowledge-base/10_LOG.md
  • knowledge-base/11_PROVENANCE.md
  • knowledge-base/12_NLP_PREPROCESS.md
  • knowledge-base/13_AUTORUN.md
  • quick-start/INIT_GUIDE.md
• All "Pending English translation" banners removed; English is now the unambiguous canonical version.

Translation impact

• Russian copies under i18n/ru/ still carry their original source_commit values. After this release, python3 scripts/check_translations.py will flag every RU file as stale (their commit lags the canonical EN sources). Re-translation is tracked via Phase 4 / future commits.
• i18n/ru/ files remain functional (they continue to describe the same system); the divergence is structural (English wording vs Russian wording), not semantic.

[0.6.0] - 2026-05-16

Phase 4 baseline: upgrade tooling, troubleshooting docs, contributor guide, drift hook.

Added

• scripts/kb_upgrade.py — upgrade tool for deployed KBs:
  • Diffs deployed scripts against the source repo
  • Detects user customizations via SHA-256 against the previous git tag
  • Writes .new sidecars for customized files instead of overwriting
  • Bumps instructions_version after a clean upgrade
  • Modes: --dry-run, --diff, --force
• docs/UPGRADING.md — step-by-step upgrade guide for KB owners
• docs/TROUBLESHOOTING.md — common installation, pipeline, lint, watcher, repomix, reflection, and translation problems with solutions
• docs/CONTRIBUTING.md — contributor onboarding
• scripts/hooks/pre-commit-translation-drift.sh — opt-in git pre-commit hook that warns (without blocking) when canonical files are committed without matching translation updates

Translation impact

• New EN documents in docs/ (TROUBLESHOOTING, UPGRADING, CONTRIBUTING) are EN-only by design (contributor-facing). They are not part of the user-facing knowledge-base/ instructions and don't need i18n/ru/ mirrors.

[0.7.0] - 2026-05-16

Phase 4 expanded: more role templates, KB health metrics, edge-case tests, top-level scripts coverage, GitHub Actions CI, badges.

Added

• Four new role templates with full placement_examples (artifacts + quickstart + don't-drop):
  • examples/founder.yml (startup founder)
  • examples/researcher.yml (researcher / analyst)
  • examples/product-manager.yml
  • examples/content-creator.yml
• kb_lint --metrics flag — KB health metrics:
  • Lifecycle distribution (permanent / evolving / temporal / unset)
  • Pages per subfolder
  • Importance avg, median, distribution
  • Freshness buckets (≤30d / 30–90d / >90d / no date)
  • Wikilink density, orphan rate, entity coverage
  • Routing depth, insight ratio, annotation overflow count
• Test fixtures in knowledge-base/scripts/tests/fixtures/:
  • with-frontmatter.md, no-frontmatter.md, empty.md, broken-frontmatter.md, sample.txt
• Edge-case tests test_edge_cases.py (16 tests) covering empty files, Unicode filenames, broken frontmatter, idempotency under churn, scaling to 50 pages
• Tests for scripts/check_translations.py (10 tests)
• Tests for scripts/kb_upgrade.py (14 tests)
• pyproject.toml with pytest config (importlib mode, dual testpaths)
• .github/workflows/ci.yml with five jobs:
  • Tests (Python 3.11, 3.12, macOS-latest)
  • Coverage upload to Codecov
  • kb_doctor self-test
  • Shell-script syntax + shellcheck
  • YAML config and role-example validation
  • Translation drift report

Changed

• README.md — updated badges (tests, coverage, version), expanded role templates table (7 roles), added module 00 and 14 to instruction list

Test stats

• 108 tests passing (was 67)
• ~69% overall coverage; kb_common 91%, kb_lint 84%, kb_reflect 70%

Translation impact

• The new role templates and updated README.md need re-translation in i18n/ru/. check_translations.py will mark them stale automatically.

[0.7.1] - 2026-05-16

Patch release: Russian translations synchronized with current canonical EN.

Added

• i18n/ru/knowledge-base/00_OVERVIEW.md — Russian translation of the deployment-overview module
• i18n/ru/knowledge-base/14_INITIAL_POPULATION.md — Russian translation of the initial-population module (was an EN copy in the previous release)
• scripts/sync_translations.py — utility that bumps source_commit, source_version, and translated_at in translated frontmatter after manual re-verification (modes: --to-head, --to-version, --lang, --files, --dry-run)

Changed

• i18n/ru/README.md — refreshed to mirror canonical EN: badges, 7-role table, 15 instruction modules (00–14), kb_lint --metrics mention, kb_upgrade.py callout, links to docs/TRANSLATING.md and docs/CONTRIBUTING.md
• i18n/ru/knowledge-base/README.md — reading order now includes modules 00 and 14; pointer to canonical EN added
• All 18 RU translation files bumped to source_version: 0.7.0 and translated_at: 2026-05-16
• docs/TRANSLATING.md — documented the post-commit workflow: git commit → sync_translations.py --to-head → check_translations.py --update-status

Translation impact

• After this release lands and is committed, check_translations.py will flag the 18 RU files as stale (their source_commit will lag the new HEAD). Run python3 scripts/sync_translations.py --to-head --lang ru immediately after committing to reset the markers.

[0.8.0] - 2026-05-16

Adds deterministic generation of DATA_PLACEMENT_EXAMPLES.md and a custom-role workflow.

Added

• scripts/kb_populate.py — pure-templating yaml → markdown generator:
  • Reads placement_examples from examples/<role>.yml
  • Emits a structured DATA_PLACEMENT_EXAMPLES.md with all sections (intro, generic table, role-specific quick map, detailed examples, quickstart, do-not-drop, footer)
  • --from <path> for custom role files
  • --create-samples writes raw/_samples/<artifact>.example.md placeholders
  • --dry-run, --json, --output <name> flags
  • 0 tokens to generate; deterministic across all AI agents
• knowledge-base/templates/role.yml.template — scaffold for the custom-role flow (Path B). The agent now must save examples/<slug>.yml before invoking kb_populate.py.
• scripts/tests/test_kb_populate.py — 25 tests covering YAML parsing, markdown rendering, edge cases, all CLI flags, and a parametrized integration test that runs against every shipped role template.

Changed

• knowledge-base/14_INITIAL_POPULATION.md rewritten around the two-path workflow:
  • Path A (built-in role): kb_populate.py --role <role> directly
  • Path B (custom role): build YAML from role.yml.template first, then Path A
  • Added an explicit "AI review pass" step where the agent appends a ## Project notes section with project-specific tips (not capturable in YAML)
  • Updated the agent checklist to enforce the new ordering
• knowledge-base/02_INIT.md Phase 3 section now points to the populate-script-first workflow

Test stats

• 133 tests passing (was 108)
• New parametrized tests catch placement_examples regressions in any role template before they ship

Translation impact

• i18n/ru/knowledge-base/14_INITIAL_POPULATION.md and i18n/ru/knowledge-base/02_INIT.md need re-translation. After committing, run:

  python3 scripts/sync_translations.py --to-head --lang ru
  python3 scripts/check_translations.py --update-status
  

[0.8.1] - 2026-05-16

Added

• examples/fiction-writer.yml — role template for fiction writers. Pillars:
  • Craft theory entities (story structure, dialogue, POV, scene mechanics, show-don't-tell)
  • Influences entity (excerpts from books the writer loves, dissected for why they work — the AI's voice anchor)
  • Genre conventions entity (tropes, comp titles, reader expectations)
  • Project hierarchy (current WIP with outline, characters, worldbuilding subfolders)
  • 11-artifact placement_examples covering excerpts, drafts, outlines, character bibles, worldbuilding, beta feedback, voice memos, genre research, queries, anti-patterns
  • ai_assistant_tasks focused on analytical critique — pacing audits, voice checks, POV slip detection, show-vs-tell passes, structural mapping
• README.md and i18n/ru/README.md — added the new role to the templates table

Test stats

• 134 tests passing (one new parametrized integration entry for fiction-writer)

[0.8.2] - 2026-05-16

Changed

• CI workflows disabled. The repo is currently local-only with no public GitHub Actions instance, so the workflow file was renamed to prevent unintended runs:
  • .github/workflows/ci.yml → .github/workflows/ci.yml.disabled
  • GitHub Actions only picks up *.yml/*.yaml files in this directory, so the .disabled suffix turns the workflow into a no-op.
  • Re-enabling is a single git mv away (see .github/workflows/README.md).
• Added .github/workflows/README.md with explicit "do not enable without the project owner's request" note + re-enable instructions.

Added

• .github/workflows/README.md — disabled-state documentation
• examples/fiction-writer.yml (released earlier this day, expanded with the long-form-book artifact entry)
• kb_ingest.py heuristic _looks_like_long_book — detects PDF/EPUB/DOCX with ≥25 000 words and adds a "long-form reference book" warning block to the generated review package, advising the agent to keep prose in assets/ and write a takeaways note instead. Metadata now records long_book_hint: true/false.
• 03_PIPELINE.md — new section "Handling long reference materials" with the asset-as-reference + notes-as-knowledge pattern and an explicit decision tree.
• 8 new tests in test_kb_ingest.py for the long-book detection and review- package wording.

Removed

• i18n/ru/knowledge-base/templates/ — empty leftover folder. Templates, scripts, and shell wrappers are language-agnostic and live only in the canonical knowledge-base/.

Test stats

• 142 tests passing (was 134)

Translation impact

• After committing, all 18 RU files become stale (HEAD moved). Run:

  python3 scripts/sync_translations.py --to-head --lang ru
  python3 scripts/check_translations.py --update-status
  

[0.9.0] - 2026-05-16

Real-world feedback round: deployed on macOS, found six UX issues, fixed all of them.

Added

• knowledge-base/shell/install.sh — flattens the setup/ payload into the project root and removes the empty setup/ folder afterward. Eliminates the nested-knowledge-base/ problem reported from a real macOS deployment.
• macOS double-click launchers (Finder-friendly, no terminal navigation needed):
  • shell/watcher-start.command — opens a Terminal window with the watcher running
  • shell/watcher-stop.command — stops a daemonized watcher
  • shell/reindex.command — one-shot manual reindex
• Windows equivalents: shell/watcher-start.bat, shell/reindex.bat
• templates/START_HERE.md.template — auto-generated post-deployment cheat sheet that reminds the user to start every new chat with the "Read AGENTS.md…" line
• !review command — drains the review queue, processes each item with explicit reporting, honors the long-form-book hint, asks the user one question per item when input is required
• !populate command — regenerates DATA_PLACEMENT_EXAMPLES.md after a role YAML edit

Changed

• Deployment flow rewritten in 02_INIT.md and 00_OVERVIEW.md:
  • User copies knowledge-base/ to their project as setup/
  • Agent runs bash setup/shell/install.sh to flatten the payload into the project root
  • No more nested knowledge-base/ inside the project — files live at the project root
• kb_populate.py moved from scripts/ → knowledge-base/scripts/ (it's a deployment-time tool, not a maintainer tool); path resolution updated to find examples/ in either layout
• templates/AGENTS.md.template — explicit "Read AGENTS.md first" workflow callout at the top, plus !review and !populate rows in the commands table
• 04_REVIEW.md — added a full !review command spec with long-form-book guard, defer-to-user pattern, batched questions
• Main README.md and i18n/ru/README.md — new "Running the watcher" section with platform-specific instructions, updated commands table with !review/!populate, project-root layout reflecting the new deployment flow, explicit "Read AGENTS.md first" warning

Test stats

• 142 tests still passing (the kb_populate move didn't change behavior)

Translation impact

• After committing, the 18 RU files become stale (HEAD moved). Run:

  python3 scripts/sync_translations.py --to-head --lang ru
  python3 scripts/check_translations.py --update-status
  

[0.9.1] - 2026-05-17

Hotfix: deployment flow direction reversed. The 0.9.0 install.sh was pre-deploy (flatten before agent works), which mixed source instructions with build artifacts and left the project messy. Replaced with finalize.sh which runs at the very end.

Changed

• knowledge-base/shell/install.sh → knowledge-base/shell/finalize.sh
  • Old behavior (pre-deploy): flatten setup/ into project root, agent builds the base on top → mixed instructions and artifacts at the root, setup/ deleted before deployment finished.
  • New behavior (post-deploy): agent builds the base inside <project>/knowledge-base/ while setup/ stays put as the source. After kb_doctor.py passes and START_HERE.md is generated, the agent runs bash setup/shell/finalize.sh which:
    • Validates required files (AGENTS.md, kb.config.yml, scripts/kb_ingest.py)
    • Refuses if anything in knowledge-base/ would overwrite a file at the project root (without --force)
    • Promotes every entry to the project root
    • Removes the empty knowledge-base/
    • Removes setup/ (use --keep-setup to retain)
  • Flags: --dry-run, --keep-setup, --force, --kb-dir <path>
• 02_INIT.md, 00_OVERVIEW.md, 14_INITIAL_POPULATION.md updated to reflect the corrected flow:
  • Agent works inside knowledge-base/ during deployment
  • --kb-root knowledge-base for kb_populate.py invocations during build
  • finalize.sh is the final step in Phase 3

Added

• knowledge-base/scripts/tests/test_finalize_sh.py — 8 tests covering happy path, refusal scenarios (missing kb, missing required files, conflicts), --keep-setup, --dry-run, --force, idempotency

Test stats

• 150 tests passing (was 142)

Translation impact

• After committing, the 18 RU files become stale (HEAD moved). Run:

  python3 scripts/sync_translations.py --to-head --lang ru
  python3 scripts/check_translations.py --update-status
  

[0.9.2] - 2026-05-17

Hotfix: shell wrappers and macOS launchers were resolving paths relative to their own folder instead of the project root, so they couldn't find scripts/kb_watch.py. Reported from a real macOS deployment.

Fixed

• shell/watcher.sh, shell/reindex.sh, shell/lint.sh, shell/doctor.sh now resolve the project root (parent of shell/) and cd there before locating scripts/. Previously they did cd "$(dirname "$0")" and ended up inside shell/, where scripts/kb_*.py does not exist.
• shell/watcher-start.command, shell/watcher-stop.command, shell/reindex.command (macOS launchers) now detect the project root whether the launcher lives in shell/ or at the project root, with a clear error message if scripts/ cannot be found.
• shell/watcher-start.bat, shell/reindex.bat (Windows launchers) — same resolution logic.
• All wrappers now report where they looked when a script is missing, pointing at the resolved project root path.

Added

• knowledge-base/scripts/tests/test_shell_wrappers.py — 8 regression tests covering:
  • Wrappers called from project root
  • Wrappers called via absolute path from arbitrary CWD
  • Helpful error message when scripts/ is missing
  • .command launcher located in shell/ (default)
  • .command launcher copied to project root (alternative layout)

Test stats

• 158 tests passing (was 150)

Translation impact

• After committing, the 18 RU files become stale (HEAD moved). Run:

  python3 scripts/sync_translations.py --to-head --lang ru
  python3 scripts/check_translations.py --update-status
  

[0.9.3] - 2026-05-17

Polish: cleaner final layout. *.sh wrappers live only in shell/, while *.command / *.bat launchers are promoted to the project root by finalize.sh so they're discoverable for double-clicking.

Changed

• shell/finalize.sh now performs two extra cleanup steps after promoting knowledge-base/ to the project root:
  1. Promotes launchers — moves every *.command and *.bat from shell/ up to the project root so Finder / Explorer can launch them. Skips files that already exist at the root (preserves user customizations).
  2. Deduplicates *.sh — if a *.sh exists both at the root and in shell/ and they are byte-identical, removes the root copy. Different content (user customization) is preserved untouched.
• Documentation updated to reflect the cleaner layout:
  • README.md, i18n/ru/README.md — Linux invocation now shown as ./shell/watcher.sh etc.; project-root tree no longer lists *.sh at the top level.
  • 00_OVERVIEW.md, 02_INIT.md — same.
  • templates/START_HERE.md.template — Linux instructions point at ./shell/..., troubleshooting tips updated.

Added

• 4 new regression tests in test_finalize_sh.py:
  • test_finalize_promotes_command_launchers_to_root
  • test_finalize_keeps_sh_in_shell_only
  • test_finalize_preserves_root_sh_when_different
  • test_finalize_promotion_does_not_overwrite_existing_root_launcher

Test stats

• 162 tests passing (was 158)