npx claudepluginhub javimontano/jm-adk-alfaThis skill uses the workspace's default tool permissions.
agents/guardian.mdagents/lead.mdagents/specialist.mdagents/support.mdevals/evals.jsonknowledge/body-of-knowledge.mdknowledge/knowledge-graph.mdprompts/meta.mdprompts/primary.mdprompts/variations/deep.mdprompts/variations/quick.mdreferences/domain-knowledge.mdtemplates/output.docx.mdtemplates/output.htmlSearches, retrieves, and installs Agent Skills from prompts.chat registry using MCP tools like search_skills and get_skill. Activates for finding skills, browsing catalogs, or extending Claude.
Generate structured prompt files for multi-agent ecosystems. Covers 9 prompt types from system prompts to committee deliberation to fallback recovery. [EXPLICIT]
{{var}}) are filled by the orchestrator/prompt-creator meta_prompt data-analyst
/prompt-creator handoff_prompt customer-onboarding
/prompt-creator committee_deliberation # interview mode
Parse $1 as prompt type, $2 as owning agent ID. If missing, ask. [EXPLICIT]
Read agents/$2/agent.md — ground prompt in agent identity, tone, and constraintsGlob agents/$2/prompts/*.md — avoid duplicatesRead references/prompt-types-spec.md if available| # | Type | File Pattern | Purpose | Complexity |
|---|---|---|---|---|
| 1 | agent_system_prompt | agent.md | Full constitution | → Use /agent-constitution-creator |
| 2 | meta_prompt | prompts/meta-{topic}.md | Behavioral instruction for ONE aspect | Low |
| 3 | system_user_pair | prompts/pair-{scenario}.md | Reusable system+user template | Low |
| 4 | workflow_step_prompt | Inline in skill.yaml | Step-level LLM instruction | → Use /workflow-creator |
| 5 | handoff_prompt | prompts/handoff.md | Task transfer protocol | Medium |
| 6 | committee_deliberation | prompts/deliberation.md | Independent multi-agent evaluation | High |
| 7 | committee_synthesis | prompts/synthesis.md | Merge multiple agent responses | High |
| 8 | validation_prompt | prompts/validation.md | Quality validation of outputs | Medium |
| 9 | fallback_prompt | prompts/fallback.md | Recovery when primary fails | Medium |
Types 1 and 4 redirect to specialized skills — this skill handles types 2, 3, 5-9. [EXPLICIT]
Write to agents/{agentId}/prompts/{filename}.md: [EXPLICIT]
---
type: "{promptType}"
owningAgent: "{agentId}"
sourceAgentMd: "agents/{agentId}/agent.md"
version: "1.0.0"
---
# {Title}
{Content per type-specific rules below}
# Reasoning Meta-Prompt
You are {{agent.name}}. Your role: {{agent.role}}. [EXPLICIT]
## Framework
1. {Step 1 of reasoning process}
2. {Step 2}
## Constraints
- {Hard limit 1}
- {Hard limit 2}
## System and ## User sections{{placeholders}}Bad handoff_prompt:
Hand off the task to the next agent. [EXPLICIT]
Good handoff_prompt:
## Handoff Protocol: {{source_agent}} → {{target_agent}}
### Context to Pass
- Task ID: {{task_id}}
- Current state: {{state_summary}}
- Completed steps: {{completed_steps}}
- Pending decision: {{decision_needed}}
### Context to Omit
- Internal reasoning chains
- Failed approaches and why they failed
- Intermediate calculations
### Success Criteria
The handoff is complete when {{target_agent}} confirms: [EXPLICIT]
- [ ] Context received and understood
- [ ] Can proceed without clarification
- [ ] Estimated completion: {{time_estimate}}
{{placeholders}} are named descriptively (not {{x}})Author: Javier Montaño | Last updated: March 12, 2026
| Scenario | Handling |
|---|---|
| Empty or minimal input | Request clarification before proceeding |
| Conflicting requirements | Flag conflicts explicitly, propose resolution |
| Out-of-scope request | Redirect to appropriate skill or escalate |