assemble-context

name: assemble-context description: | Assemble a task-specific context brief by searching across multiple sources (Slack, Notion, git repos, Temporal docs, meeting notes). Runs as a subagent to avoid burning tokens in the main conversation. Use when starting a non-trivial task that benefits from gathering relevant context from enterprise sources first. Triggers: "assemble context", "gather context", "research context for", "context brief", "seed prompt"

Assemble Context

Given a task description, launch a subagent that searches across available sources, triages results for relevance, and writes a structured context brief to a file. The main conversation stays clean; context arrives as a file you can reference with @.

Workflow

1. Ask the user for the task description if not already provided.

2. Ask which sources to search (default: all). Available sources:

   • Slack — via agent-slack CLI
   • Notion — via user-notion MCP (notion-search, notion-query-meeting-notes, notion-fetch)
   • Git repos — at ~/src/temporal-all/repos and any other relevant repos
   • Temporal docs — via user-temporal-docs MCP (requires auth first)
   • Meeting notes — via notion-query-meeting-notes
   • Meeting transcripts — via hyprnote-search CLI (see ~/.agents/skills/hyprnote/SKILL.md)
   • Temporal Cloud operational details - via oncall repo

3. Launch a generalPurpose subagent using the Task tool. Pass it a prompt constructed from the template in subagent-prompt.md, substituting the task description and selected sources.

4. The subagent writes its output to ~/context-briefs/<slug>.md where <slug> is a short kebab-case name derived from the task.

5. Tell the user the file path so they can @-reference it.

Notes

• Use multiple subagents in parallel if the task has clearly separable research areas.
• The subagent should prefer breadth over depth: surface many potentially-relevant items with brief summaries, rather than deeply reading a few.
• Token budget guidance for the brief: aim for 15-30K tokens of assembled context. Prioritize by estimated relevance. Summarize long items; include short items verbatim.