Turn requests into planned work
Turn a support escalation into planned work
Write the issue first, plan from the issue, and approve only the sub-issues the evidence supports, so the agent fixes the bug you reported and nothing else.
Run your morning without missing anything
Find what actually needs you today across What’s next, the Sessions panel, and the command palette, and hand the first item off before standup.
Make a technical decision and keep the reasoning
Research an open question read-only, record the decision as a document, and attach it to the issue, so the team can audit it six months later.
Work a long document without losing what it says
Interrogate a spec, brief, or RFC in its own context chat, get answers with quoted lines you can check, and turn the gaps into issues without leaving the page.
Hand work to agents
Hand an issue to a coding agent without a round trip
Put seven things in the issue and Cursor, GitHub Copilot, or Bright’s own Engineer can execute it unattended and come back with a pull request.
Encode a recurring review as a skill and an agent
Write the release checklist once as a skill with an output contract, give an agent a refusal, and stop re-explaining your standards.
Worked examples end to end
Build a documentation portal from a product brief
Turn a product brief into a published docs portal by giving each Bright surface one job and keeping the product decisions yours.
Investigate a suspected page failure
Tell a broken page from a client-rendered one with a read-only investigation that stays inside a step cap and knows when to stop.
Run a GitHub and Notion launch review
Audit launch readiness across GitHub and Notion without writing anything, and end with one report whose every claim you can open at its source.
Create and cross-link GitHub and Notion artifacts
Create a GitHub repository and a Notion brief, cross-link them, and hand off only what you verified at the provider, never a success message.
Good to know
- Every playbook ends with a “What good looks like” checklist. Use it as the acceptance test for the run.
- Writes to the workspace or to a connected provider pause for a tool approval. Read the card, then allow once.
- The page failure and launch review playbooks start in Ask mode and write nothing, but connector reads still show one approval card each.
- The technical decision playbook starts in Ask mode and switches to Agent only to write the decision document.