Run subagents in forked contexts with isolated prompts.
npx skills add https://github.com/CaptainCrouton89/crouton-kit --skill a-name
Please help me install this Agent Skill. Skill: a name Source: https://github.com/CaptainCrouton89/crouton-kit/tree/main/plugins/.claude/skills/example Command: npx skills add https://github.com/CaptainCrouton89/crouton-kit --skill a-name
This Skill demonstrates the use of forked contexts to run subagents in isolated environments, preventing prompt leakage between invocations.
Invoke the a name skill in a forked context with your arguments, e.g., a name arg1 arg2.
High-intent search queries and answers about installing and using this skill.
Forked-context execution isolates subagents by running each invocation in a fresh context with its own system prompt. This approach prevents prompt leakage and cross-run contamination, keeping arguments and prompts contained within the current run.
Prompt isolation keeps prompts and arguments contained within a single agent run, preventing cross-run contamination. You need it for testing agent prompts, sandboxed experimentation, and scenario-specific agent behavior without leaking context between invocations.
You pass arguments to a subagent in a forked context by invoking the skill with your desired parameters, such as appending arguments like arg1 and arg2 to the invocation. The arguments stay contained within that isolated run.
Yes, forked-context execution maintains a separate system prompt for each invocation. Running a subagent in a fresh context ensures that the system prompt and arguments are isolated and specific to that individual run.
Forked contexts for subagent workflows prevent cross-run contamination but are primarily designed for testing agent prompts and sandboxed experimentation. They isolate prompts per run, meaning shared state or persistent context across invocations is not maintained.
TechyMTConfigure slash commands for inline or forked subagent execution.
etylsarinCoordinate three reviewer agents to vote on a shared document with a single click.
youdotcom-ossRun controlled web search agent evaluations across CLI agents in Docker containers.
onelee85Coordinate subagents and persistent state for multi-session debugging workflows.
