Discipline plugins for Claude Code — prompt engineering, agent dispatch, writing, and search.
User: "Research X with oberweb"
→ Identify search dimensions (docs, tutorials, discussions)
→ Dispatch parallel sonnet search agents
→ Synthesize results with user's model + source URLs
User: "Create a skill for code review"
→ INTAKE: Problem, triggers, 5+ use rule
→ DESIGN: Single vs router, freedom level
→ BUILD: SKILL.md with anti-rationalization table
→ TEST: Baseline (RED) → Compliance (GREEN) → Loopholes (REFACTOR)
→ SHIP: Pre-flight checklist, package
User: "Review this agent prompt"
→ Load review-prompt.md checklist
→ Structure, efficiency, security audit
→ Verdict table with fixes
User: "This sounds too robotic, fix it"
→ PURGE: Strip aidiolect, em-dashes, hollow openers, hedges
→ STRUCTURE: Fix burstiness, vary rhythm
→ STRUNK: Active voice, concrete language, omit needless words
→ VOICE: Add specificity, positions, gaps, contractions
User: "Take a screenshot and analyze it"
→ Capture screen (full, active window, or named window)
→ Dispatch haiku analyzer
→ Return summary with insights
| Skill | Purpose | Example |
|---|---|---|
| oberprompt | Prompt engineering principles | "use oberprompt to fix this flaky agent" |
| oberagent | Agent dispatch validation | "use oberagent before dispatching" |
| oberweb | Multi-dimensional web search | "use oberweb to research X" |
| obercreate | Skill creation AND review | "create a skill for X" / "review this prompt" |
| oberscribe | Human-sounding writing (Strunk + AI pattern detection) | "this sounds robotic, fix it" / "oberscribe this README" |
| obershot | Screenshot capture and analysis (full, active, or named window) | "take a screenshot and analyze it" / "screenshot of Firefox" |
oberweb (standalone)
│
└── Dispatch parallel sonnet search agents
│ (extract and distill precise info, no summaries)
└── Synthesize with user's model + source URLs
obercreate (standalone)
│
├── CREATE: INTAKE → DESIGN → BUILD → TEST → SHIP
│ │
│ └── Pressure testing with fresh subagents
│
└── REVIEW: Load checklist → Audit → Verdict table
oberagent (invoked before any agent dispatch)
│
└── oberprompt
│
└── Prompt engineering
(constraints, validation)
oberscribe (standalone)
│
└── PURGE → STRUCTURE → STRUNK → VOICE
│
└── Hard rules (em-dash ban, aidiolect kill list)
+ Craft (specificity, rhythm, voice, gaps)
obershot (standalone)
│
└── Capture (full / active / --mode window --name "App")
│
└── Dispatch haiku analyzer → Return summary
# Add marketplace (if not already added)
/plugin marketplace add ryanthedev/rtd-claude-inn
# Install plugin
/plugin install oberskills@rtd
# Update to latest
/plugin update oberskills@rtdCurrent version: 1.24.0
Each skill displays its version at runtime by reading from .claude-plugin/plugin.json (e.g., obershot v1.20.0).
MIT