Change Log

All notable changes to the BASH Consultants repository will be documented in this file.

Check Keep a Changelog for recommendations on how to structure this file.

[Unreleased]

Platform transformation: activated the site’s dormant AI capabilities, wired the dead conversion paths, added CI quality gates, and expanded the content and services catalog.

Added

• Privacy policy page — privacy.md (/privacy/), an honest, plain-English privacy policy grounded in what the site actually collects: PostHog analytics (opt-in, consent-gated, session recording off), the mailto-based contact/newsletter forms, the currently-dormant AI chat assistant, and GitHub Pages hosting. Covers cookies, the consent banner, service providers, data retention, and rights under the GDPR, CCPA/CPRA, and Colorado Privacy Act — without claiming any certification. Linked from the contact page; content_lint.py now also gates root privacy.md. Doubles as a prerequisite for the LinkedIn Community Management API application.
• PostHog analytics integration (server-side API) — corrected the site’s client key in _config.yml (it was the zer0-mistakes theme’s bundled demo key, so the site’s analytics never reached our own project — project 504877 had ingested zero events) and enabled IP anonymization on the project. Added scripts/features/posthog/ (stdlib Python) which uses the private POSTHOG_API_KEY personal key to verify the site is wired to the correct project and its privacy settings haven’t drifted, print settings, and post deploy annotate markers; .env.example documents the key. Event retention is plan-controlled on PostHog Cloud and cannot be set through the API.
• LinkedIn company-page publishing — a governed draft→approve→publish pipeline for the BASH LinkedIn page (urn:li:organization:64517157). The stdlib-only Python publisher scripts/features/linkedin/ posts article link-shares of blog posts and standalone text updates via the Posts API, uploads the post’s preview image as the card thumbnail via the Images API, and records each share in an idempotency ledger (.github/linkedin-log.json) so nothing double-posts; --dry-run renders the exact payload with no API calls, and it reuses content_lint.py’s frontmatter reader and banned-phrase list as a mechanical brand guard. A /linkedin-draft command + linkedin-share skill draft on-brand commentary into drafts/linkedin/ (status: pending) and open a PR — the human merge is the approval. Two workflows: linkedin-publish.yml (manual dispatch + merge-triggered from-drafts, commits the ledger back with [skip ci]) and linkedin-token-health.yml (weekly warning before the 60-day token expiry). The company page is wired into _config.yml, _data/entity/info.yml, and the JSON-LD sameAs; .env.example documents the server-side-only LINKEDIN_* credentials.
• Weekly content review — automation + a full first-pass review — a governed AI agent that moves the site’s content forward by one unit each week: it reviews the corpus and opens a PR that either expands an existing article or writes a new one, never pushing to main. Ships the content-curator charter (.claude/agents/content-curator.md), the content-review.yml workflow (Thursdays), and scripts/content_inventory.py — a deterministic thin/stale shortlist that seeds the weekly run (the same script-first pattern as the preacher). The first run was an exhaustive parallel review of all ~40 content pages; its top verified findings are applied below.
• New post: “When the numbers say it’s time to leave QuickBooks” (pages/_posts/corp/) — the CFO-level decision post for the practice’s flagship QuickBooks-to-ERP line, which the review found named everywhere but never given a decision piece: the threshold signals, the honest cost-of-inaction math, total cost in ranges, and vendor-neutral targets, citing Oracle NetSuite and Microsoft Dynamics 365 Business Central primary docs. Cross-linked with the ERP service and the outgrowing-QuickBooks guide.
• The preacher — automated “practice what we preach” enforcement — a governed AI agent that runs every Saturday (.github/workflows/preacher.yml) and holds the repo to the doctrine the site sells (the adapted principles DFF/KIS/DRY/REnO/MVP/COLAB/AIPD plus deterministic-first, governed, build-in-the-open). Two modes: when it finds violations it opens ONE GitHub issue; when the repo is clean it mechanizes one recurring AI-review burden into a deterministic check and opens a PR — so the deterministic floor rises and future enforcement costs less AI. Ships with scripts/doctrine_check.py (an extensible check registry; seed check DRY-CONTACT catches the exact contact-info duplication that prompted this), the preacher subagent charter (.claude/agents/preacher.md), and the canon/runbook (docs/the-preacher.md). It never pushes to main — issues and PRs only. Also reconciled docs/automation.md (added the preacher + pr-to-upstream, marked the retired Azure workflow).
• Upstream sync automation — .github/workflows/pr-to-upstream.yml opens (and keeps) a pull request from bamr87:main up to the repository this was forked from (amr-bash/bashconsultants) on every push to main. Idempotent (won’t duplicate an open PR, skips when already in sync) and safe (skips cleanly when the UPSTREAM_PR_TOKEN secret is absent). Landed alongside a keep-ours reconcile merge so bamr87/main cleanly supersedes upstream’s older post pass, keeping those PRs conflict-free.
• Native Claude context framework — turned the repo’s AI tooling into a first-class Claude Code framework and made the repository its own reference implementation of the governed-AI model the site sells. Added a root CLAUDE.md (the canonical Claude entry point, routing into the existing AGENTS.md/instructions/prompts plus the operating rules) and .claude/README.md (the map of the layer). Codified the brand as a single source of truth (.github/instructions/brand.instructions.md — name/etymology, mission, voice, doctrine, colors, logo). Added four skills (.claude/skills/{content-editorial,toolkit-doc,wikilinks,brand}), two subagents (.claude/agents/{jekyll-build-validator,brand-guardian}), and three commands (.claude/commands/{lint-content,new-toolkit-doc,brand-check}) — the agents the practice describes at /ai-operations/ now exist as files. Documented the whole pattern on-site as a partner-track guide, pages/_toolkit/claude-context-framework.md (/tools/partners/claude-context-framework/), with a mermaid architecture diagram.
• The BASH toolkit collection — turned the single /tools/ page into a two-track docs/training library (toolkit collection under pages/_toolkit/, served at /tools/business/… and /tools/partners/…). The business track is 10 plain-English DIY guides for SMB owners (IT foundations, adopting AI, cloud migration, outgrowing QuickBooks, faster close, dashboards, integrations, IT roadmap, security baseline, backups); the partner track is 10 advanced technical references for consultants (engagement method, deterministic-first doctrine, AI-native practice, production AI/RAG, cloud landing zones, ERP architecture, record-to-report, modern data stack, integration architecture, security architecture), several with mermaid architecture diagrams. Data-driven hub + track landings, a toolkit sidebar nav, and a “Toolkit” top-nav dropdown; the old internal-tooling content moved into the partner-track “Running an AI-native consulting practice” doc. Cross-linked with Obsidian wiki-links; every doc cites a verified-live primary source.
• Two new service pages — pages/_services/ai.md (/services/ai/) and pages/_services/managed-it.md (/services/managed-it/), completing the eight-service catalog; the flagship AI offering previously existed only as a hub section with the landing card mis-linked to /services/dev/.
• Case studies collection — pages/_case_studies/ (six anonymized engagement snapshots + index), registered in _config.yml, surfaced in the landing proof band.
• /tools/ and /ai-operations/ — root pages presenting the firm’s own AI toolchain (prompt playbook, VS Code orchestrator, image pipeline, governance stack) as client-relevant proof, with _data/playbook.yml auto-generated from the prompt/instruction files by scripts/generate_playbook_data.py.
• Four pillar posts (2026-07-06) — an SMB AI acceptable-use policy, cyber-insurance renewal questions, MCP for the back office, and the AI audit trail for finance; plus the promoted bashos post from drafts/.
• AI chat proxy — api/ Azure Functions app implementing the theme’s chat-proxy contract (server-held Anthropic key, model pinned server-side, per-IP rate limiting keyed on the trusted proxy hop, body-size and token caps); staticwebapp.config.json pins the managed-functions runtime to Node 20.
• CI and automation — .github/workflows/build-validate.yml (builds both production stacks + content lint on every push/PR), site-health.yml (nightly), and content-gardener.yml (weekly draft PRs); scripts/content_lint.py mechanically enforces the editorial contract.
• AI-search layer — /llms.txt, site-wide + per-type JSON-LD (Article/Service) in _includes/structured-data.html, and a wired /search/ page over search.json.
• _data/taxonomy.yml, docs/preview-images.md, docs/automation.md, and .devcontainer/ — codified the section voice profiles, preview-image and automation runbooks, and a container dev environment.

Changed

• Canonical domain set to bash-365.com. The repo referenced bashconsultants.com (now a separate mirror — the bash-amr fork) as the site URL, so canonical tags, the sitemap, structured data (_data/entity/info.yml), the logo link, the LinkedIn publisher’s share URLs, and the privacy policy all pointed at a stale mirror whose /privacy/ 404s. Switched _config.yml url, the entity data, the LinkedIn base URL default, docs, and the README/privacy references to the live main domain bash-365.com. Email addresses (@bashconsultants.com) are unchanged.
• Analytics privacy hardening (live-validation follow-up) — after end-to-end testing on the deployed site, closed two gaps the validation surfaced: disabled PostHog IP geolocation ($geoip_disable in the posthog.html override) so no city/region/postal is derived from a visitor’s IP (the project’s anonymize_ips only stops the raw IP being stored, not GeoIP enrichment), and made PostHog persistence in-memory until analytics consent is stored, so no PostHog cookie or localStorage is written before opt-in. Corrected the privacy policy’s IP wording to match. The same validation confirmed prior opt-in, GPC/Do-Not-Track suppression, zero Google tracking, and IP-not-stored were all already working.
• Removed Google tracking; PostHog is now the sole analytics. Disabled Google Analytics 4 (commented google_analytics in _config.yml) and neutralized the theme’s hardcoded Google Tag Manager container with a local no-op override at _includes/analytics/google-tag-manager-head.html. Hardened PostHog to true prior opt-in via a local _includes/analytics/posthog.html override: it initializes opted-out (opt_out_capturing_by_default: true) so it captures nothing before consent, and it honors Global Privacy Control (GPC) alongside Do Not Track — under either signal PostHog does not load. Closes the consent-enforcement gaps the privacy-policy review surfaced.
• Service pages expanded and reconciled (first content-review pass) — pages/_services/erp.md gained a “Moving off QuickBooks to a mid-market ERP” section (naming Oracle NetSuite and Microsoft Dynamics 365 Business Central as destinations) and a “What it changes” outcomes block in honest categories, and its QAD versions are now labeled legacy. The services hub (index.md) description now names all eight services (adding AI and managed IT), the three divergent financial-systems platform lists were reconciled (dropping the unsupported Anaplan outlier), the QAD/legacy-ERP versions are labeled “versions we’ve worked on” with a link to QAD’s current Adaptive ERP line, and the IT-strategy roadmap horizon was aligned so the hub no longer contradicts the strategy page.
• Contact info deduplicated to a single source of truth — the identical email/phone/LinkedIn/website block that was copy-pasted at the foot of all eight service pages is now one shared include, _includes/contact-methods.html, rendered from _data/entity/info.yml (contact-email, contact-phone), the socials list, and the canonical info.url. The remaining hardcoded contact values on contact.md, about.md, pages/_services/index.md, and the founder page’s hero + contact tiles now read from that same data (personal email from site.email), so the phone number, address, and domain each live in exactly one place. Changing a value once updates it everywhere.
• Founder page redesigned (/about/amr/) — added a designed profile hero (avatar, identity, social links, consultation call-to-action), a stat band mixing career numbers with live GitHub stats for @bamr87 (30 repos, 19 stars, 17 followers, coding in public since 2015), a row of open-source project cards (it-journey, zer0-mistakes — the theme this site runs on — bamr87.github.io, aieo), and technical-skills and strengths as badge groups; the full CV detail is preserved. The GitHub stats are deterministic-first: real static values render without JavaScript and a small client-side script only ever updates them from the GitHub API, never blanks them. The Education, Professional experience, and Project record sections were then restyled to match — education as cards, experience as a blue timeline (with each role’s dates, and Navistar’s three roles and nested location lists preserved), the project record as per-employer panels with project-count pills, and the contact block as a grid of icon tiles with tel:/mailto: links — with all CV detail intact. Accent color uses the UI blue per the brand palette — the brand-guardian agent caught an initial use of logo-teal as the UI accent, which the brand SSOT forbids.
• Registered the Claude layer and standardized the brand name — AGENTS.md and .github/copilot-instructions.md now point to CLAUDE.md, the brand SSOT, and the .claude/ primitives; the file-scoped instruction map routes brand-bearing files to brand.instructions.md. Aligned the BASH expansion to the canonical Bourne Again Solutions Hero (the form that actually spells the acronym) in llms.txt, matching README.md.
• Landing page rebuilt as data-driven sections (_layouts/landing.html + index.md frontmatter + _data/entity/*.yml): hero, eight service cards, six industry pain-point cards, process, an “AI-augmented practice” section, proof band, and FAQ; removed the Thrasymachus epigraph.
• Hero particle background fixed — the #particles-js element was a Bootstrap .container (capped at 1140px, centered), so the particle canvas rendered as a bounded box; it now fills the hero edge-to-edge with the content layered above.
• Obsidian wiki-links — internal cross-references across service, post, and case-study pages now use pipeless [[Page Title]] wiki-links (Obsidian-native, vault-portable), resolved to real links by the theme’s client-side resolver over wiki-index.json. Pipeless is required: the server-side plugin can’t run under GitHub Pages safe mode, and kramdown mangles the | of aliased wiki-links into broken tables. Root-page and CTA links (/tools/, /ai-operations/, /contact/) stay markdown.
• powered_by refreshed — corrected to the real toolchain (Ruby 3.3, Jekyll 3.10, zer0-mistakes 1.26.0, Bootstrap 5.3.3, Claude, OpenAI gpt-image, Python 3.13, Docker, GitHub Pages/Actions, Azure Static Web Apps); dropped the removed Algolia and the stale HTML/CSS/JS/jQuery filler.
• Conversion forms wired — the contact and newsletter forms now POST to config-driven endpoints (site.forms.*) with graceful mailto: fallbacks; previously both had no handler.
• Theme pinned — remote_theme pinned to bamr87/zer0-mistakes@v1.26.0 and Gemfile.azure raised to the matching ~> 1.26.0 (with jekyll-include-cache); a registered aqua skin replaces the invalid dark value; theme_color hex values quoted so they stop parsing as null.
• Editorial debt burn-down — sentence-cased service-page titles, removed unbacked “robust/seamless/scalable”, stripped trailing periods from service descriptions, retitled the ERP section off its self-labeling name, and realigned article-review.prompt.md and the reviewer agent to the house rules they enforce.
• content_statistics_generator.rb made Jekyll-3/4 compatible (Azure stack was crashing on Theme#root_dir).

Removed

• Azure Static Web Apps deploy workflow — retired .github/workflows/azure-static-web-apps-proud-pond-06dc10c1e.yml. It was failing the CI gate independently of any change (its deployment token is unset, and its Oryx build can’t run bundle), while GitHub Pages is the primary and working deploy target. The Azure build is still validated by build-validate.yml (Gemfile.azure + _config.azure.yml), and staticwebapp.config.json / api/ remain for a future reconfiguration if Azure hosting is revived.
• Stale artifacts — the drifted scripts/lib/preview_generator.py, clipboard-era scripts/{vscode-integration.sh,demo-docs-generation.js,test-docs-prompt.js}, the broken Rakefile, and the stale Algolia config block; rewrote the root README.md from stale marketing copy (wrong bash-365.com contact, 3-of-8 services) into a repo-facing developer doc.

[1.5.0] - 2026-06-21

Added

• pages/_posts/muses/2025-01-24-bash-consulting-breaking.md — New satirical press-release post (“If a press release about ethical capitalism wrote itself”) covering ESG accounting for Denver SMBs; includes a straight practical section on what’s buildable on QuickBooks today.
• .github/prompts/article-write.prompt.md — New /article-write agent prompt for drafting new bashconsultants.com posts end-to-end.

Changed

• pages/_posts/muses/2025-08-12-innovation-paradox.md — Frontmatter updated (author, description, keywords, lastmod, tags normalized); content rewritten for voice and SMB focus.
• pages/_posts/tech/2025-08-12-ai-integration-reality.md — Frontmatter updated; content rewritten with honest SMB-first framing on AI implementation.
• pages/_posts/tech/2025-11-19-prompts-are-the-new-command-line.md — Frontmatter normalized (author, lastmod, categories/tags as YAML lists); content already updated to SMB voice.
• Site-wide content review — Rewrote about.md, contact.md, and all service pages (pages/_services/{cloud,erp,data,dev,fintech,strat}.md) to comply with content-style.instructions.md: removed banned marketing phrases, switched to sentence-case headings and the 8-section service-page structure, defined acronyms on first use, and added inline images with descriptive alt text.
• pages/_services/index.md — Rebuilt as a card-based hub linking to all six real service pages (it previously listed generic categories that did not match the site’s actual services).
• _data/entity/info.yml — Cleaned banned phrases from the About / Founder / Mission copy that feeds the landing page.
• Blog posts — Added a next-step call-to-action to posts that lacked one; normalized preview: paths to the /assets/images/previews/ convention; bumped lastmod on edited posts.
• All 11 blog posts (pages/_posts/{corp,erp,muses,tech}/*.md) — Full editorial pass aligning every post to content-style.instructions.md and posts.instructions.md: heavy rewrite to the business-outcome-first voice, acronyms defined on first use, banned marketing phrases removed, sentence-case headings, and a single next-step CTA plus one outbound primary-source link in every post. Tightened titles to sentence case ≤ 60 chars and description to 120–155 chars.
• Creative voice — show, don’t announce — Removed raw genre/device labels (“satire”, “parody”, “reality show”, “Broadway-style musical number”, “exposé”) from reader-facing text so each post enacts its device (the press release, the show conceit, the song, the Frankenstein metaphor) instead of narrating it.
• pages/_posts/tech/2025-08-12-ai-integration-reality.md — Published the post (was draft: true with duplicated author/description keys); kept the stronger “AI for small business: what the pilots don’t tell you” version and rewrote the body as an SMB-first AI-adoption guide.
• Post frontmatter normalized site-wide — author: "Amr Abdel-Motaleb" everywhere (replacing “BASH Consulting Team” / “bamr87”); lowercase categories; flow-style tags / keywords lists; duplicate keys removed; lastmod set to 2026-06-21.
• Preview images — Regenerated 9 AI preview banners (gpt-image-2, retro pixel-art) to match the rewritten titles and content; standardized all post preview: paths to the /images/previews/<slug>.png short form (auto-normalized by the theme).

Removed

• pages/_posts/tech/2025-01-24-bash-consulting-breaking.md — Deleted from tech/; replaced by the muses rewrite above.
• Invented case studies and unverifiable metrics — Removed fabricated results (“40% cost reduction”, “99.9% uptime”, “300% revenue increase”, “30+ ERP projects”, “over 15 years”) and overreaching capability claims (e.g., AI/ML clinical decision support, fraud detection) from the service pages, per the “never invent customer wins” rule.

Fixed

• Broken images — Renamed assets/images/red-rocks.JPG → red-rocks.jpg so the lowercase reference resolves on case-sensitive production builds; repointed the missing office-594119.png og_image to office.jpg in _config.yml and pages/_about/site/_config.yml; repointed the missing bamr-avatar.png avatar to Amr-Headshot_v2.jpg.
• Contact-info consistency — Standardized all customer pages to bashconsultants.com / info@bashconsultants.com / (720) 352-4641; replaced the stale bash-365.com details on about.md and contact.md; fixed the contact-email: info@consultants.com typo in _config.yml that broke the landing page’s “Email Now” button.

Added

• .github/instructions/ — Six file-scoped instruction files for AI agents:
  • content-style.instructions.md (applyTo pages/**, index.md, about.md, contact.md) — shared editorial voice, audience profile (Denver SMBs 5–200 employees), target-industry table, banned phrases, SEO + accessibility rules
  • posts.instructions.md (applyTo pages/_posts/**/*.md) — blog post structure + subfolder routing (corp/erp/muses/tech)
  • services.instructions.md (applyTo pages/_services/**/*.md) — service-page conversion structure + industry-fit naming
  • jekyll-theme.instructions.md (applyTo theme override paths) — local override rules for the remote theme
  • extension.instructions.md (applyTo extension/**) — VS Code Prompt Orchestrator dev rules
  • prompts.instructions.md (applyTo .github/{prompts,instructions}/**) — schema + slash-hard body style
• AGENTS.md — Cross-tool entry point for AI coding agents (Copilot, Codex, Cursor, Aider, Claude Code). Points to canonical guidance.
• .github/prompts/commit-publish.prompt.md — Standard commit + push + deploy workflow for the direct-to-main publishing model.
• Canonical frontmatter reference — .github/FRONTMATTER.md documents the required schema for .prompt.md and .instructions.md files.

Changed

• .github/copilot-instructions.md — Rewritten to reflect the actual dual-purpose repo (Jekyll site + extension/ sub-project). Previous content described only the extension and was stale.
• .github/prompts/ frontmatter normalized — All .prompt.md files updated to canonical mode: agent + description + date + lastmod schema. Body content already concise (≤100 lines per file); no further compression needed.

Removed

• .github/prompts/docs.prompt.md — superseded by documentation.prompt.md.

[1.4.0] - 2026-04-06

Changed

• Theme Upgrade: Updated zer0-mistakes theme from v0.21.6 to v0.22.10
  • Gemfile.azure: Updated jekyll-theme-zer0 from ~> 0.21.2 to ~> 0.22.10
  • pages/_about/site/_config.yml: Updated remote_theme pin from v0.21.6 to v0.22.10
  • _data/README.md and _data/navigation/README.md: Updated version references to v0.22+
  • Notable theme changes since v0.21.6:
    • Admin layout and configuration dashboards (v0.22.10)
    • Cross-platform installation compatibility fixes (v0.22.8)
    • Hero layout and scroll animation stabilization (v0.22.3)
    • Various documentation and CI improvements

[1.3.0] - 2026-03-31

Changed

• Theme Upgrade: Updated zer0-mistakes theme from v0.18.0 to v0.21.6
  • Gemfile: Updated jekyll-theme-zer0 from ~> 0.18.0 to ~> 0.21.2 (latest on RubyGems)
  • Gemfile.azure: Added version pin ~> 0.21.2
  • _config.yml: Pinned remote_theme to bamr87/zer0-mistakes@v0.21.6
• Layout Standardization: Aligned post and section layouts with theme v0.20.3+ conventions
  • Changed default post layout from journals to article in _config.yml defaults
  • Updated 9 posts with explicit layout: journals or layout: blog to layout: article
  • Updated 6 category/index pages to use layout: default (theme’s section layout uses /news/ URLs)
  • Renamed category pages with date prefixes for proper Jekyll _posts processing
• New Features Enabled:
  • Added search.json for GitHub Pages compatible search (theme v0.20.2 feature)
  • Added mermaid diagram configuration for client-side rendering
  • Added jekyll-include-cache plugin for improved build performance
  • Updated vendor exclusion paths to support assets/vendor/ publishing
• Configuration Updates:
  • Updated Bootstrap version reference from 5.2.0 to 5.3.3 in powered_by
  • Updated version references in _data/README.md and _data/navigation/README.md to v0.21+
  • Synced pages/_about/site/_config.yml remote_theme pin and post layout default

Fixed

• 404 Page: Fixed broken /blog navigation link to use correct /posts path
• Navigation: Removed broken links to non-existent pages (/docs/, /about/org/, /about/portfolio/)
• Navigation: Fixed /home/ links to / across navigation data files
• Blog Index: Fixed hardcoded post URLs using slug format to use date-based format
• Category Pages: Added missing permalink to muses category page

[1.2.0] - 2026-01-24

Changed

• Development Port Standardization: Updated all development ports from 4002 to 4042 for consistency
  • Dockerfile: Changed EXPOSE and CMD to use port 4042
  • docker-compose.yml: Updated port mapping to 4042:4042
  • .vscode/tasks.json: Updated Jekyll local serve port to 4042
  • .vscode/launch.json: Updated all debug configuration URLs to localhost:4042
• VS Code Development Environment: Complete overhaul of development configurations
  • .vscode/launch.json: Added 6 new debug configurations:
    • 🐳 Debug BASH Consultants (Docker) - Primary Docker debugging
    • 🔗 Attach to Running Jekyll (Docker) - Attach to running container
    • 🔄 Docker Rebuild & Debug - Force rebuild with debugging
    • 🖥️ Debug BASH Consultants (Local) - Local development debugging
    • 📱 Mobile Debug (Docker) - Mobile responsive testing
    • 🎯 Performance Debug (Docker) - Performance profiling
  • .vscode/tasks.json: Added 12 comprehensive tasks:
    • Docker operations (up, stop, down, rebuild, logs)
    • Jekyll serve/build (Docker and Local)
    • Maintenance tasks (clean up, update dependencies)
• Theme Version: Updated jekyll-theme-zer0 to v0.18.0
  • Gemfile: Updated gem version constraint to ~> 0.18.0
  • _config.yml: Simplified remote_theme to bamr87/zer0-mistakes (uses latest)
• Gemfile Cleanup: Simplified dependencies
  • Removed redundant jekyll_plugins group (already included in github-pages gem)
  • Removed pinned versions for ffi and webrick (use latest compatible)
  • Removed commonmarker pin (resolved upstream)

Added

• .gitignore: Added logs/ directory for debug trace logs

Fixed

• Dockerfile: Added newline at end of file

[1.1.0] - 2026-01-24

Changed

• Navigation Data Structure (_data/navigation/): Updated to zer0-mistakes theme v0.17+ format
  • Migrated from sublinks to children property for nested navigation items
  • Added Bootstrap Icons (bi-* prefix) to all navigation items
  • Added consistent Home links to sidebar navigation files
  • Added trailing slashes to all URLs for consistency

Added

• New Navigation Files (_data/navigation/)
  • home.yml - Homepage quick navigation with icons
  • services.yml - Services sidebar navigation
  • README.md - Navigation schema documentation with migration guide

Fixed

• Fixed HTML entity & in about.yml to proper & character
• Fixed inconsistent URL trailing slashes across navigation files

[1.0.0] - 2026-01-24

Added

• Prompt Orchestrator VS Code Extension (extension/): New VS Code extension for orchestrating AI agent workflows using structured prompts
  • Discovers and loads prompt templates from .github/prompts/ directory
  • Execute prompts via Command Palette or sidebar tree view
  • Integration with VS Code Chat API and Language Model API
  • Commands: refreshPrompts, executePrompt, review, refactor, test, docs, debug
  • Configuration settings for prompts directory and auto-refresh
• Prompt Engineering Toolkit (.github/prompts/)
  • docs.prompt.md - Technical documentation generation
  • Additional prompts for requirements analysis, system design, code implementation, test generation, code refactoring, and debugging
• Automation Scripts (scripts/)
  • routine-maintenance.sh - Prepare prompt workflows for VS Code Chat
  • demo-docs-generation.js - Demonstrate documentation prompt generation
  • test-docs-prompt.js - Standalone test script for prompt loading
  • vscode-integration.sh - VS Code CLI integration helper
• Blog Posts (pages/_posts/)
  • “Prompts: The New Command Line” - Exploring the paradigm shift in development
  • “From Prompts to Pipelines: Agentic Workflows in VS Code” - Extension design and agentic workflows

Changed

• Repository Structure: Reorganized to separate Jekyll website from VS Code extension
  • Jekyll site configs remain at root level
  • Extension code moved to dedicated extension/ directory
  • Updated .gitignore for new structure
• VS Code Workspace Configuration (.vscode/)
  • Updated launch.json for Jekyll-focused development
  • Updated tasks.json for Jekyll build tasks
  • Updated settings.json for Liquid/Jekyll file associations
  • Updated extensions.json with Jekyll-relevant recommendations

• Root package.json: Clean Jekyll-only build configuration for Azure Static Web Apps

• README.md: Added Tools and Resources section documenting prompt engineering toolkit

Removed

• Extension files from root directory (moved to extension/)
  • tsconfig.json, esbuild.js, eslint.config.mjs (now in extension/)
  • src/, out/, dist/ directories (now in extension/)
  • .vscode-test.mjs, .vscodeignore, vsc-extension-quickstart.md, EXTENSION-README.md