Hermes Skills — Production-Grade Agent Skills for Flutter, DevOps & Agent Operations

A curated collection of battle-tested agent skills forged in production Flutter projects (CarSah, Hermex Android, Azdal). Every pattern, pitfall, and protocol earned through real debugging sessions — 48 documented patterns and counting.

Two AI agents — an implementer and an independent auditor — review each other's work with no shared runtime and no API between them. The git repo is the mailbox: every message is a commit, both sides poll a marker for changes, and state accumulates in one shared record. Full protocol: agent-repo-handoff-loop — 112 KB, field-hardened.

Transport Who talks to what

Cadence How work advances and state accumulates

Exchange One round-trip, in order

Each diagram also ships as .svg (vector) and self-contained .html (source) under docs/diagrams/.

The skills above are free and open-source (Apache 2.0) — that's the framework.

Premium is the automation that builds and verifies the entire Spec Pack for you:

Buy once, use forever. Instant digital delivery — no subscription, no account required.

• 🔗 Purchase: spec-pack-premium on Lemon Squeezy
• 🛒 Also listed: MCP Market

npx skills add jahfaliabdulrahman-dev/hermes-skills

npx skills add jahfaliabdulrahman-dev/hermes-skills --skill specification-writing
npx skills add jahfaliabdulrahman-dev/hermes-skills --skill flutter-soul-stewardship
npx skills add jahfaliabdulrahman-dev/hermes-skills --skill what-if-analysis
npx skills add jahfaliabdulrahman-dev/hermes-skills --skill translation-fidelity
npx skills add jahfaliabdulrahman-dev/hermes-skills --skill flutter-isar-clean-arch-setup
npx skills add jahfaliabdulrahman-dev/hermes-skills --skill flutter-android-build-system
npx skills add jahfaliabdulrahman-dev/hermes-skills --skill flutter-design-anti-patterns
npx skills add jahfaliabdulrahman-dev/hermes-skills --skill flutter-lessons-patterns
npx skills add jahfaliabdulrahman-dev/hermes-skills --skill flutter-patterns
npx skills add jahfaliabdulrahman-dev/hermes-skills --skill flutter-isar-testing
npx skills add jahfaliabdulrahman-dev/hermes-skills --skill flutter-sdk-changelog
npx skills add jahfaliabdulrahman-dev/hermes-skills --skill flutter-ios-build-system
npx skills add jahfaliabdulrahman-dev/hermes-skills --skill device-screen-verification
npx skills add jahfaliabdulrahman-dev/hermes-skills --skill agent-repo-handoff-loop
npx skills add jahfaliabdulrahman-dev/hermes-skills --skill mcp-bridge-patterns
npx skills add jahfaliabdulrahman-dev/hermes-skills --skill repo-front-door
npx skills add jahfaliabdulrahman-dev/hermes-skills --skill supabase-fullstack
npx skills add jahfaliabdulrahman-dev/hermes-skills --skill github-project-audit
npx skills add jahfaliabdulrahman-dev/hermes-skills --skill skill-ecosystem-sync

Every skill in this repository was born from a production failure that became a permanent lesson:

• LL-010 (Live-Device Verification): 5 critical bugs found on real device AFTER flutter analyze clean + 34/34 tests passing + 2 AI auditors signing APPROVE. → flutter-lessons-patterns now enforces live-device verification as the final gate before any DONE declaration.
• LL-024 (ClassNotFoundException): namespace in build.gradle.kts didn't match MainActivity.kt package. → flutter-android-build-system now gates every build with 5 preflight checks.
• LL-011 (Disabled Button Colors): Material silently substitutes its default palette when onPressed: null — custom colors lost without error. → Same skill now enforces explicit disabledBackgroundColor/disabledForegroundColor.
• The cutoff-bridge with its own cutoff: an upstream skill sold as "bridging the AI training cutoff" shipped a Flutter matrix that stopped at 3.27 / 3.29 — 6 stable minors and ~18 months behind — plus two wrong Dart mappings, because its CI validated frontmatter, link targets and prose style but never a single version number. → flutter-sdk-changelog generates every number from Google's release index and the flutter/flutter stable tree, and ships a freshness gate that exits 1 the day stable moves ahead (proven with 5 tamper tests, not just a green run).
• SOUL Quality Conundrum: EPIC-001 failed in 27 minutes because 2-3 line SOUL.md files gave agents no identity. After rewriting SOULs to 150-581 lines, EPIC-002 succeeded. → flutter-soul-stewardship is the real lever — not rule-file scaffolding, but strong agent identity.

We don't ship "best practices." We ship scar tissue.

These skills are free and open-source (Apache 2.0). If they save you hours of debugging, consider supporting the work:

• 💖 GitHub Sponsors — recurring support
• ☕ Buy Me a Coffee — coming soon

Every contribution funds more battle-tested patterns, device verification, and new skills.

Each skill follows the skills.sh format:

skills/<name>/
├── SKILL.md          # YAML frontmatter + procedural knowledge
├── references/       # Deep-dive technical references
├── scripts/          # Automation scripts
└── assets/           # Templates, images, configuration

To contribute a skill to this collection:

1. Fork this repo
2. Create your skill directory under skills/
3. Write SKILL.md with proper YAML frontmatter (name, description, version)
4. Include references and scripts as needed
5. Submit a PR

Apache 2.0 — see LICENSE.

These skills are open-source and free to use, modify, and distribute. Attribution appreciated.

Eng. Abdulrahman Jahfali — energy-sector operations engineer turned AI/Flutter architect. Every skill here was forged in real battles: production apps, real devices, and agents held to evidence instead of promises.

• GitHub: @jahfaliabdulrahman-dev
• Projects: CarSah, Hermex Android, Azdal, RQS V3.1, Kronos

_{Built with battle scars. Verified on device. Every number generated, not remembered.}