A complete skill library for running a sovereign AI operating system from the command line. Built by Paul Desai as part of the Active Mirror project — a sovereign AI infrastructure running entirely on a Mac Mini M4 with zero cloud dependencies.
144 skills. 184 backing scripts. 83 automated agents. One machine.
Every skill is a /command for Claude Code (Anthropic's CLI agent). Drop them in ~/.claude/commands/ and they become invocable with /skill-name.
These aren't toy prompts. Each skill is a job profile — a complete workflow with steps, verification, and output format. They represent every job a sovereign AI system needs to do, organized by domain:
| Skill | What it does |
|---|---|
/forensic |
Full 10-layer deep system scan (services, network, kernel, integrity, infra, deps, vault, security, continuity, drift) |
/benchmark |
Quantitative performance benchmarks tracked over time |
/regression |
Detect performance regressions against baselines |
/wire-check |
Verify all AI components are connected end-to-end |
/calibrate |
Self-benchmarking with grade (A-F), trend, and recommendations |
/drift |
Detect configuration drift across all sources of truth |
/promote |
Promote script to managed service (LaunchAgent + registry) in one step |
/retire |
Gracefully disable a service (stop, archive, update registry) |
/device |
Device mesh status dashboard |
/cert |
SSL certificate expiry monitor |
/inventory |
Complete system inventory (scripts, agents, skills, ports, repos, models) |
/cost |
API spend and resource tracking |
/dependency |
Package and vulnerability audit (brew, pip, npm) |
/health |
Service health check |
/status |
Quick status (bus, inbox, git, queues) |
/pulse |
Self-healing infrastructure check |
/smoke |
End-to-end smoke tests |
/services |
Service management |
/deploy |
Deploy LaunchAgents |
/killswitch |
Emergency stop all agents |
/emergency |
Emergency triage protocol |
/network |
Network diagnostics |
| Skill | What it does |
|---|---|
/evolve |
Self-improvement cycle: calibrate, identify weaknesses, fix |
/breakthrough |
Track and generate novel capabilities |
/capability |
System capability matrix with coverage scores |
/prove |
Generate verifiable proof that a capability works |
/experiment |
Controlled A/B experiments on system behavior |
/adversarial |
Red team self-testing |
/dream |
Run the Dream Engine (autonomous self-improvement) |
/reflect |
Meta-analysis of own performance |
/council |
Multi-perspective review (architect, critic, operator) |
| Skill | What it does |
|---|---|
/commit |
Smart commit with semantic message |
/ship |
Commit, push, and mark complete |
/pr |
Create pull request |
/release |
Create release |
/scaffold |
Scaffold new project |
/test |
Run tests |
/review |
Review changes |
/factory |
Visual orchestration demo |
/swarm |
Multi-agent swarm coordination |
/spec |
Draft specification document |
| Skill | What it does |
|---|---|
/draft |
Draft content for publishing |
/paper |
Generate research paper from system capabilities |
/storytell |
Narrative content generator |
/hook |
Generate viral video hooks |
/headline |
Generate headlines and titles |
/cta |
Generate call-to-action variants |
/idea |
Generate content ideas |
/carousel |
Social media carousel |
/quote-card |
Quotable image cards |
/behind-scenes |
Behind-the-scenes content |
/evergreen |
Create evergreen content |
/comparison |
Create comparison content |
/challenge |
Coding challenge content |
/newsletter |
Weekly newsletter |
/community-post |
YouTube community post |
/thread |
Turn build log into social thread |
/note |
Quick capture to inbox |
/brief |
Sovereign briefing generator |
| Skill | What it does |
|---|---|
/daily-video |
Compile daily build log video |
/weekly-video |
Weekly recap video |
/shorts |
Generate YouTube Shorts from recordings |
/clip |
AI-powered highlight clipper |
/timelapse |
Coding timelapse |
/video-essay |
AI-narrated video essay |
/podcast |
Extract podcast audio |
/tutorial |
Extract tutorial from build log |
/stack-video |
Generate "My Stack" video |
/endscreen |
Optimize end screens |
| Skill | What it does |
|---|---|
/auto-publish |
Syndicate content to all platforms |
/distribution |
Full distribution pipeline |
/launch |
Content launch sequence |
/drip |
Drip content publisher |
/beacon |
Manage blog |
/seo-syndication |
SEO and content syndication |
/repurpose |
Repurpose content across formats |
/remix |
Remix old content |
/schedule |
Optimal content schedule |
/content-calendar |
Content calendar management |
/sitemap |
Regenerate sitemaps |
/schema |
Add structured data / Schema.org |
/meta |
Optimize page meta tags |
/youtube-seo |
AI-native YouTube SEO |
/og-image |
Generate Open Graph images |
| Skill | What it does |
|---|---|
/analytics |
Cross-platform analytics |
/growth |
Growth metrics dashboard |
/velocity |
Content velocity tracker |
/funnel |
Content funnel analysis |
/engagement |
Engagement booster |
/ab-test |
A/B test video titles and thumbnails |
/viral-audit |
Viral readiness audit |
/trending |
Find trending topics |
/niche-down |
Niche analysis and positioning |
/audience |
Audience research and personas |
/compete |
Analyze competitor channels |
/hashtag |
Research optimal hashtags |
| Skill | What it does |
|---|---|
/find |
Smart vault search |
/read |
Follow a guided read path |
/audit |
Vault health audit |
/triage |
Auto-triage inbox |
/compress |
Identify compression candidates |
/link |
Vault link management |
/tag |
Tag management |
/glossary |
Glossary management |
/migrate |
Move notes between vault layers |
/focus |
Set current project focus |
/inbox-zero |
Clear all pending items |
/research |
Deep research |
| Skill | What it does |
|---|---|
/incident |
Post-mortem generator from failure data |
/replay |
Reconstruct sessions from witness chain |
/contract |
Work contracts for digital coworkers |
/onboard |
Generate context pack for new agents |
/handoff |
Write handoff for next agent |
/pickup |
Continue from last handoff |
/standup |
Morning standup |
/overnight |
Queue background agent tasks |
/signal |
Send notification via sovereign notification service |
/morning-push |
Send morning briefing to devices |
/phone-pull |
Pull and triage files from mobile |
/sync |
Synchronize state across all clients |
/bus |
Memory bus operations |
/history |
Session history |
| Skill | What it does |
|---|---|
/brand |
Brand consistency checker |
/collab |
Find collaboration opportunities |
/backlink |
Backlink strategy |
/milestone |
Celebrate and share milestones |
/recap |
Weekly/monthly content recap |
/changelog |
Generate changelog |
/thumbnail |
AI-generate video thumbnails |
/demo |
Mobile app demo orchestration |
/snapshot |
System snapshot |
/optimize |
Full content optimization sweep |
/guardian |
Mirror Guardian integrity check |
| Skill | What it does |
|---|---|
/decay |
Check staleness thresholds |
/reaper |
Staleness reaper (auto-archive) |
/backup |
Vault backup operations |
/logs |
View service logs |
/ffmpeg-reaper |
Kill zombie FFmpeg processes |
/cognitive |
Cognitive dashboard |
/research-agents |
Run intelligence sweep |
/release |
Create release |
These skills are the interface layer of a sovereign AI stack:
You (human)
|
v
Claude Code + 141 Skills (this repo)
|
v
Cognitive Kernel (28 primitives, temperature-as-architecture)
|-- 15 deterministic (temp 0.0): identity, search, verify
|-- 6 creative (temp 0.7): dream, compose, simulate
|-- 7 blended (temp 0.3): conscience, learn, anticipate
|
v
14 Substrates (conscience, witness, dream, calibrate, ...)
|
v
Self-Healing Infrastructure (MirrorPulse)
|-- Detect -> Diagnose -> Heal -> Learn -> Report
|-- Self-learning playbook (evolves fix patterns)
|
v
184 Scripts + 83 LaunchAgents + 5 Ollama Models
|
v
Mac Mini M4 (24GB, sovereign, no cloud required)
- Temperature-as-Architecture: Each kernel primitive declares its mode (deterministic/creative/blended). Temperature isn't a parameter — it's a structural decision.
- Conscience Loop:
DREAM -> CALIBRATE -> CONSCIENCE -> INJECT -> ACT -> WITNESS— the system evaluates its own behavior and nudges itself. - Witness Chain: Cryptographic SHA256 hash-chain providing provenance for every AI session.
- Compaction Survival: Distill checkpoints that survive context window compression.
- Self-Healing Playbook: MirrorPulse learns which fixes work and evolves its repair patterns.
# Clone
git clone https://github.com/MirrorDNA-Reflection-Protocol/sovereign-ai-skills.git
# Copy skills to Claude Code
cp sovereign-ai-skills/skills/*.md ~/.claude/commands/
# Now you can use /forensic, /benchmark, /evolve, etc. in Claude CodeMost skills assume a MirrorDNA-style directory structure at ~/.mirrordna/. Adapt paths for your setup. The skill patterns are transferable to any sovereign AI infrastructure.
Paul Desai — Building sovereign AI infrastructure at Active Mirror. 8 published papers. 57 repos. Zero cloud dependencies.
- activemirror.ai — Main site
- beacon.activemirror.ai — Blog
- Publications
MIT