Brandon Martin
f80d95a869
## Summary - ✨ Add skill system with 3 skills (cloudflare-vpc-services, code-reviewer, commit-helper) and skill skeleton template - ✨ Add 8 specialized agents for ArgoCD, Braintrust, GitHub Actions, Fumadocs, and Helm - ✨ Add 4 new slash commands (/agent-intent-from-docs, /create-skill, /triage, /wrapup-skillup) - 📝 Add comprehensive plugin system documentation (plugins.md, plugins-reference.md, plugin-marketplaces.md, agent-skills.md) - 🔧 Update install_extensions.py to support skill discovery and installation - 🔧 Add skill_forced_eval hook and update settings.json ## Test plan - [x] Verify skills are discovered by running `uv run install_extensions.py list --type skill` - [x] Test `/create-skill` command creates proper skill structure - [x] Verify new agents appear in `/agents` interface - [x] Confirm new commands work via `/help` - [x] Review documentation renders correctly
6.3 KiB
Create Agent from Documentation with Intent
You will create a new specialized agent by analyzing documentation from the provided URLs, guided by a specific intent or purpose.
Arguments: $ARGUMENTS
Important: The first argument is the intent/description of what the agent should do. All remaining arguments are documentation URLs to analyze.
Your Task
Follow these steps to create the agent:
Step 1: Parse Arguments
Extract from $ARGUMENTS:
- Intent: The first argument (everything in quotes or up to the first URL)
- Documentation URLs: All remaining arguments
The intent describes the specific purpose or use case for this agent (e.g., "Create a migration assistant agent from incident.io and opsgenie to betterstack").
Step 2: Analyze Documentation
If multiple URLs are provided, launch multiple technical-researcher agents in parallel (one per URL). If only one URL is provided, use a single researcher.
For parallel execution with multiple URLs:
- Launch all technical-researcher agents simultaneously in a single message with multiple Task tool calls
- Each agent should analyze its assigned URL independently
- Combine all research findings before proceeding to Step 3
Each researcher should investigate with the intent in mind:
- What technology/framework/library is being documented
- Key concepts, APIs, methods, and patterns relevant to the stated intent
- Best practices and conventions
- Common workflows and use cases that align with the intent
- Important implementation details that an expert should know
- Migration paths, comparison features, or integration points if relevant to the intent
- How this technology differs from alternatives
Request a comprehensive summary suitable for creating a specialized agent focused on the stated intent.
Step 3: Determine Agent Creation Strategy
After gathering all research findings, analyze whether to create one or multiple agents:
Create MULTIPLE agents when URLs represent distinct verticals/technologies:
- Different core APIs or services (e.g., Cloudflare Workers vs D1 vs R2)
- Separate deployment models or runtime environments
- Distinct problem domains within a platform
- Independent tools/products that happen to share a brand
Create a SINGLE agent when URLs cover the same technology:
- Different aspects of one framework (e.g., Next.js routing + data fetching + deployment)
- Progressive documentation depth on one topic
- Multiple guides for the same library/tool
- When the intent requires unified knowledge across all URLs (e.g., migration agents need to understand both source and destination)
Examples:
- ✅ Single agent: Intent="migration from X to Y" + docs for X + docs for Y → 1 migration specialist
- ✅ Multiple agents:
workers.cloudflare.com/docs+developers.cloudflare.com/d1+developers.cloudflare.com/r2→ 3 specialized agents - ❌ Single agent:
nextjs.org/docs/routing+nextjs.org/docs/api-routes+nextjs.org/docs/deployment→ 1 comprehensive agent
Based on this analysis and the stated intent, proceed to either Step 4A (single agent) or Step 4B (multiple agents).
Step 4A: Generate Single Consolidated Agent
Use when: All URLs document the same technology/framework OR the intent requires unified knowledge across all URLs (like migration, integration, or comparison tasks).
Use the meta-agent to create one specialized agent based on all research findings.
Provide the meta-agent with:
- The stated intent as the primary goal of this agent
- Complete research findings from all URLs
- Instruction to create an agent that specializes in fulfilling the stated intent
- Request to include "MUST BE USED" or "use PROACTIVELY" in the description, with clear triggers based on the intent
- Guidance on appropriate tool restrictions for this domain
- Request to incorporate patterns and examples from the documentation
- Emphasis on how the agent should leverage knowledge from all sources to fulfill the intent
The meta-agent should write the agent file to .claude/agents/specialized/<appropriate-category>/ with a descriptive filename that reflects the intent.
Step 4B: Generate Multiple Specialized Agents
Use when: URLs represent distinct verticals/technologies within a platform AND the intent doesn't require unified knowledge across them.
For each identified vertical/technology, use the meta-agent to create a separate specialized agent.
You can launch multiple meta-agent tasks in parallel (one per vertical) by including multiple Task tool calls in a single message.
For each meta-agent invocation, provide:
- How the stated intent relates to this specific vertical
- Research findings specific to that vertical's URL(s)
- Instruction to create an agent specializing in that specific vertical while supporting the overall intent
- Request to include "MUST BE USED" or "use PROACTIVELY" in the description
- Guidance on appropriate tool restrictions for this domain
- Request to incorporate patterns and examples from that vertical's documentation
- Clear scope boundaries (what this agent handles vs other related agents)
Each meta-agent should write its agent file to .claude/agents/specialized/<appropriate-category>/ with a descriptive filename that identifies the specific vertical.
Step 5: Report Results
For single agent (Step 4A): Provide a summary including:
- The stated intent and how the agent fulfills it
- Key findings from the documentation analysis
- Path to the newly created agent file
- Agent name and description
- When and how this agent should be used (with specific triggers based on the intent)
- Suggested test scenarios to validate the agent against the stated intent
For multiple agents (Step 4B): Provide a summary including:
- The stated intent and how the agent ecosystem fulfills it
- Overview of how the platform was divided into verticals
- For each created agent:
- Path to the agent file
- Agent name and description
- What vertical/technology it covers
- How it supports the overall intent
- When and how it should be used
- Suggested test scenarios that would trigger each agent
- Explanation of how the agents complement each other to fulfill the stated intent