Gather local documents and repository state into a .wm/dive_context.md file.
npx skills add https://github.com/open-horizon-labs/bottle --skill dive-prep
Please help me install this Agent Skill. Skill: dive-prep Source: https://github.com/open-horizon-labs/bottle/tree/main/.claude-plugin/plugins/wm/skills/dive-prep Command: npx skills add https://github.com/open-horizon-labs/bottle --skill dive-prep
Dive prep reduces ambiguity at the start of an AI session by gathering relevant context from multiple sources and surfacing a grounded context file for the agent to use.
Invoke the skill with /wm:dive-prep [intent or context]. Use the Task tool to spawn the wm:dive-prep agent: Task(subagent_type: "wm:dive-prep", prompt: "<user's intent or context>") If the user provides a GitHub issue URL, include it in the prompt: Task(subagent_type: "wm:dive-prep", prompt: "Prepare dive for https://github.com/org/repo/issues/123")
High-intent search queries and answers about installing and using this skill.
Grounding context for focused AI sessions is prepared by gathering local documents, repository state, and OH context into a curated dive_context.md file that reduces ambiguity and gives the agent reliable working memory.
The best way to ground an agent before debugging is to assemble local documents, git state, and OH context into a single dive_context.md file, establishing a deterministic starting point that reduces session ambiguity.
Yes, context management allows you to spawn a deterministic agent in Claude Code by using the Task tool to invoke the wm:dive-prep agent with a specific user intent or context, bypassing direct shell commands.
You can generate a context file from a GitHub issue by invoking the dive-prep skill with the issue URL included in the prompt, which gathers repository state and outputs a grounded dive_context.md file.
Yes, dive-prep applies to Claude Code, Codex, and Open Code sessions that require reliable grounding from local documents and repository state to establish a focused working context.
A grounding file is needed for AI task spawning to reduce ambiguity at the start of a session, surfacing curated context from multiple sources so the agent operates with reliable working memory.
pradyumnasagarAutomate inter-skill data transfer and provenance capture via JSON Material Passports.
gitgoodordietryingOrchestrate coding agents via background processes in isolated workdirs.
CsHengSpawn governed subagents with scoped tool permissions via runAgent.
petrSimonidesXartAssemble per-sub-agent context bundles from evaluate-story output for deterministic spawning.
