Workflow Builder Lite

Build and execute multi-step workflows with conditional logic. Chain API calls, agent actions, and shell commands into sequences with if/else branching. Pers...

Security Scan

The skill's instructions and requirements are internally consistent with its stated purpose (building, saving, and running workflows) and it requests no extra credentials or installs — but review saved workflows and be aware that API calls and agent actions can run without per-step confirmation after a workflow is approved.

Purpose & Capability

The name/description match the SKILL.md: it builds, persists, and executes multi-step workflows including API calls, agent actions, and shell/file steps. The skill declares no environment variables, binaries, or installs — which is reasonable for an instruction-only orchestration tool.

Instruction Scope

Instructions are explicit about saving workflows to memory/workflows/{name}.md, auto-creating that directory, and executing steps. Shell commands and file writes require per-step confirmation, but API calls and agent actions are allowed to run without per-step confirmation once the workflow itself is approved — this is coherent for automation but means approved workflows may trigger external requests or agent-side actions without additional prompts.

Install Mechanism

There is no install spec and no code files; this is the lowest-risk model for static analysis (nothing is written to disk by an installer).

Credentials

The skill declares no required environment variables, credentials, or config paths. That matches the described functionality (it relies on the agent's existing tools rather than external secrets).

Persistence & Privilege

The skill persists workflows under memory/workflows/ and will auto-create that directory. always:false and no global config changes are requested. The SKILL.md asserts it won't modify other skill files, but as an instruction-only skill this is a behavioral guideline rather than an enforceable constraint.

Assessment

This skill appears to do what it claims, but take these precautions before installing or using it: - Review any saved workflow file in memory/workflows/ before running — approved workflows can trigger API calls and agent actions without additional confirmation. - Be cautious embedding secrets or credentials in workflow steps or API-call parameters; prefer using secure connectors or prompts. - For workflows that interact with external systems, test in a safe environment first. - If you want extra safety, modify the workflow or agent settings to require confirmation for API/agent-action steps as well (the SKILL.md currently only mandates confirmation for shell/file steps).

Workflow Builder Lite v2.0

Install: clawhub install workflow-builder-lite

Plan, save, and execute multi-step workflows with conditional logic.

Language

Detect from user's message language. Default: English.

How It Works

Build — Natural Language

User: "Lag en workflow: hent vær → hvis regn → send meg melding"

Agent:

Workflow: weather-alert
1. ⛅ Hent vær (web_fetch wttr.in/Oslo?format=j1)
2. 🔀 Sjekk: Er det regn?
   → Ja: 3. 📨 Send melding "Ta paraply!"
   → Nei: Stop

Godkjent? (ja/nei/rediger)

Build — Step by Step

"Legg til steg: hent data fra API X" "Legg til betingelse: hvis status != 200, stopp" "Slett steg 2"

Save — Persist to File

When user approves a workflow, save to memory/workflows/{name}.md:

# Workflow: {name}

## Steps
1. [type] description
2. [type] description
   → condition: if X then step 3, else stop

## Created
YYYY-MM-DDTHH:mm+ZZ:ZZ

## Last Run
Never

Execute — With Confirmation

When user says "kjør workflow {name}":

1. Load workflow from file
2. Show steps to user
3. Ask: "Kjør denne workflowen? Y/N"
4. Execute step by step
5. Report progress after each step

Step types and how to execute:

Type	Execution Method	Requires Confirmation
API call	web_fetch or agent built-in HTTP	No (after workflow approval)
Agent action	Built-in tools (message, browser, etc.)	No (after workflow approval)
Shell command	exec tool	Yes — show command, ask each time
File write	write tool	Yes — show content, ask each time

Shell/file steps require per-step confirmation. Show the exact command/content and wait for Y/N.

Progress

Running: weather-alert
  [✅] Step 1: Hent vær — 15°C, delvis skyet
  [⏭️] Step 2: Sjekk regn — nei
  [⬜] Step 3: (skipped)
Done. No rain today.

Quick Commands

User says	Action
"lag workflow"	Start building
"vis workflows"	List saved workflows
"kjør {name}"	Execute saved workflow
"rediger {name}"	Modify steps

Guidelines for Agent

1. Save workflows to files — memory/workflows/ for reuse
2. Confirm shell/file steps — always ask before executing
3. Report after each step — keep user informed
4. Support conditional branching — if/else based on step results
5. Auto-create memory/workflows/ if it doesn't exist

What This Skill Does NOT Do

• Does NOT execute shell commands without user confirmation
• Does NOT write files without user confirmation
• Does NOT modify MEMORY.md, HEARTBEAT.md, or other skill files
• Does NOT require external dependencies

More by TommoT2

• smart-api-connector — Connect to any REST API without code
• context-brief — Persistent context survival across sessions
• cross-check — Auto-detect and verify assumptions

Install the full suite:

clawhub install workflow-builder-lite smart-api-connector context-brief cross-check