Context documents provide essential information to AI skills about the role they should play, the agents available to them, the team they are collaborating with, and any memory that should persist across sessions. These documents help skills make better decisions and produce more relevant results.
nut context manages context documents in the current local workspace. To read context from a remote peer Coconut through the Coconut Dashboard, use nut coconut context.
type - Context type (role, agents, team, or memory)
Options:
--json - Output as JSON
Examples:
# Read role definitionnut context get role# Read agents as JSONnut context get agents --json# Read team contextnut context get team# Read memorynut context get memory
--file <path> - Path to file containing new content
--content <text> - Direct content to set
Examples:
# Update with direct contentnut context update role --content "# Role Definition\n\nAI skills should focus on code quality and testing..."# Update agents from filenut context update agents --file ./agents.md# Replace team documentnut context update team --file ./docs/team.md
Using update replaces the entire document. Use append to add content without replacing existing text.
content - Text to append (optional if using --file)
Options:
--file <path> - Path to file to append
Examples:
# Append text directlynut context append team "## New Section\n\nContent here..."# Append from filenut context append memory --file ./additional-notes.md# Add role clarificationsnut context append role "## Additional Guidelines\n\n- Always write tests\n- Follow the style guide"
Appending is useful for incrementally building up context without having to re-specify everything.
# Agents## PlannerBreaks high-level goals into ordered, reviewable tasks.## ImplementerExecutes tasks, writes code, and opens proposals for review.## ReviewerAudits proposals for correctness, security, and style before merge.
# Memory## Recent Decisions- Chose PostgreSQL over MySQL for JSONB support- Standardized on Next.js App Router## Ongoing Considerations- Evaluating queue options for background jobs- Monitoring bundle size on the marketing site
Regularly update context documents as your project evolves. Outdated information can lead to AI skills making incorrect assumptions.
Be specific and detailed
The more detail you provide, the better AI skills can understand your project and make appropriate decisions. Include coding standards, architectural patterns, and specific requirements.
Use consistent formatting
Structure your context documents with clear headings and sections. Markdown formatting makes documents easier to read and parse.
Start simple, iterate
Begin with basic context and expand over time. You don’t need perfect, comprehensive context on day one—add details as you discover what information skills need most.
