apprentice_skill

This skill helps you convert observed tasks into repeatable Python-based workflows that your agent can run forever, without writing code.
  • Python

2.5k

GitHub Stars

6

Bundled Files

2 months ago

Catalog Refreshed

4 months ago

First Indexed

Readme & install

Copy the install command, review bundled files from the catalogue, and read any extended description pulled from the listing source.

Installation

Preview and clipboard use veilstrat where the catalogue uses aiagentskills.

npx veilstrat add skill openclaw/skills --skill apprentice

  • _meta.json274 B
  • observe.py6.8 KB
  • README.md5.9 KB
  • run.py6.4 KB
  • SKILL.md8.8 KB
  • synthesize.py12.2 KB

Overview

This skill captures a task the first time you perform it and turns it into a permanent, repeatable workflow your agent can run. I watch your spoken narration and command sequence, infer intent and variables, synthesize an editable workflow, and save it locally for on-demand replay. No coding or formal specs required.

How this skill works

When you say a trigger phrase like "watch me," I enter observation mode and record the actions and narration you perform. After you stop, I analyze the log to identify ordered steps, which elements vary versus which stay constant, and produce a readable, editable workflow file plus an executable script. You review and approve the synthesis before it is saved; nothing is sent externally and execution happens only when you explicitly run a workflow.

When to use it

  • Automate repetitive shell or project setup tasks after demonstrating them once
  • Capture multi-step procedures you perform with spoken rationale and choices
  • Turn tutorial-style demonstrations into reusable automation
  • Create personal runbooks for deployments, onboarding, or weekly reports
  • Chain several learned tasks into larger automations

Best practices

  • Narrate your intent and decisions aloud while performing the task
  • Explicitly name changing items (e.g., PROJECT_NAME) during observation
  • Mark the end of the demonstration with a clear stop phrase
  • Keep each observation focused on a single cohesive workflow
  • Review and edit the synthesized workflow before approving it

Example use cases

  • Record a new-project bootstrap sequence and replay it for every repo creation
  • Demonstrate your deployment steps once and have the agent run them reliably
  • Capture your weekly reporting routine to auto-generate and send summaries
  • Teach the agent how you onboard a client so it can perform the same steps later
  • Chain project setup with notification and tracking workflows into one command

FAQ

No. All logs, synthesis, and workflow generation happen locally; I only use your active LLM session for interpretation and do not call external APIs.

How do I correct a learned workflow?

After observation you can edit the generated workflow and its variables before saving. You can also update or delete saved workflows later and re-run an observation to relearn a process.

Built by
VeilStrat
AI signals for GTM teams
© 2026 VeilStrat. All rights reserved.All systems operational