interactive
$
npx mdskill add skillmatic-ai/playbooks/interactiveDemonstrates HITL flow by asking questions, generating drafts, and requesting approval for agent validation.
- Helps validate interactive agent workflows through user input and approval steps.
- Depends on environment variables like STEP_ID, RUN_ID, and ORG_ID for execution.
- Uses free-text answers to create drafts and approval decisions to finalize reports.
- Writes results to a shared file path for review and integration.
SKILL.md
.github/skills/interactiveView on GitHub ↗
---
name: interactive
description: A test skill that demonstrates the HITL (human-in-the-loop) flow by asking a question, using the answer, and requesting approval on a draft.
---
# Interactive Skill
This skill validates the full interactive agent flow: question → answer → approval.
## Behavior
1. Ask the user a free-text question about project preferences
2. Use the answer to generate a draft document
3. Request approval on the generated draft
4. If approved, write the final report; if rejected, note the rejection
## Expected Environment
- `STEP_ID` - The step identifier within the playbook run
- `RUN_ID` - The playbook run identifier
- `ORG_ID` - The organization identifier
## Output
Writes a report to `/shared/results/{STEP_ID}/report.md` including the user's
answers and approval decision.
More from skillmatic-ai/playbooks
- account-provisionerAI-powered account provisioning with Slack welcome and Jira task creation. Uses Claude for plan generation and real API integrations. Supports approve-only approval with 2-phase checkpoint/resume.
- calendar-managerAI-powered week-1 meeting schedule generation with Google Calendar event creation. Uses Claude for schedule optimization and Google Calendar API for real events. Non-interactive, exception-only approval.
- Codebase Analysis"Analyze repository structure, tech stack, and code patterns to inform implementation planning"
- echoA test skill that echoes received input. Used for CI/CD pipeline validation and step agent container lifecycle testing.
- email-drafterAI-powered welcome email drafting with Gmail delivery. Uses Claude to generate personalized emails and sends them via Gmail API. Supports review-and-edit approval with 3-phase checkpoint/resume.
- Feasibility Report"Analyze codebase and estimate implementation effort for a proposed feature"
- JTBD Clustering"Cluster support tickets and user feedback into Jobs-to-be-Done themes"
- PRD Template v2"Write a comprehensive Product Requirements Document following the standard template"
- report-generatorAI-powered executive onboarding report compilation. Uses Claude to analyze step results and generate a narrative report. Non-interactive, auto-executes with exception-only approval.
- Sentiment Analysis"Analyze pain intensity and sentiment from customer communications"