Skip to main content

​
Basic Workflow

Reference workflow (.md) — human-readable procedure, agent follows as instructions. 
# workflows/cleanup.md
---
name: cleanup
description: Clean up temporary files
---

# Cleanup Workflow
Remove temp files and rebuild.

​
SkillFlow (executable YAML)

SkillFlow (.yaml) — deterministic, multi-step, executable pipeline. 
# workflows/data-pipeline.yaml
name: data-pipeline
description: Process data through validation, transformation, and storage
steps:
  - skill: validate-input
    prompt: "Validate the CSV data format"

  - skill: __approval_gate__
    prompt: "Data validation complete. Approve to continue?"
    channel: telegram

  - skill: transform-data
    prompt: "Transform to required schema"

  - skill: save-to-database
    prompt: "Store results"

​
Approval Gates

Steps with skill: __approval_gate__ pause execution and send an approval request via the specified channel (telegram or whatsapp). The user has 5 minutes to approve or the step times out. Useful for data-modifying operations, deployments, or any critical path requiring human sign-off.
If neither Telegram nor WhatsApp is configured, the gate auto-approves and execution continues.

​
SkillFlow Features

  • Deterministic Execution — Skills run in declared order, not LLM discretion
  • Approval Gates — Pause and require human approval via Telegram/WhatsApp
  • Multi-step — Chain multiple skills into a single automated pipeline
  • Prompt Override — Add per-step instructions with the prompt: field

Tools

Built-in and declarative tools the agent uses to act

Skills

Reusable task modules the agent learns and crystallizes over time

Hooks

Intercept, block, or modify agent behavior at every stage

Integrations

Send approval gate requests via Telegram or WhatsApp