Drop #10: Skills with Side Effects, Safe Automations & Resources Worth Bookmarking
Welcome to Drop #10. This week we are diving into one of the most powerful — and underused — patterns for building skills safely: running them in isolated subagents with context: fork. Plus, we have compiled a collection of the essential Claude Code docs and guides worth keeping bookmarked.
Tip of the Week: Build Skills with context: fork
When writing a skill with side effects — posting a message, pushing a branch, updating a ticket — add context: fork to your SKILL.md frontmatter. This runs the skill in an isolated subagent without access to your conversation history. Combine with disable-model-invocation: true so Claude only fires it when you explicitly invoke it.
---
name: send-standup
description: Post standup to Slack
context: fork
disable-model-invocation: true
---Worth Bookmarking This Week
- → Remote Control docs — continue any local Claude session from your phone or another browser
- → Official plugin marketplace — browse everything installable right now
- → Cloud scheduled tasks guide — full docs for the cloud scheduler
- → Claude Code skills reference — the definitive guide to building and sharing skills
Skills are one of the most direct ways to extend Claude — but they only work well when you have guardrails around the ones that matter. Context: fork gives you those guardrails. Use it.
Until next time.
— The ClaudeCowork.com team