Personal strategies for shipping code with Cursor, agents, and user-scoped skills and rules—not hardware setup. For machine and editor install, see ../device/macOS.md (and parity). For employer SDE2 norms, bookmarks, and team runbook stubs, see ../device/work.md. For personal life tasks (P0–P4, governance), see ../tasks/README.md.
Upstream: Notes root
- Keep one topic per file so you can link a ticket or chat to a single runbook.
- Separate how to frame and execute engineering work with AI from device shopping notes and from the personal task mirror.
| You need… | Start here |
|---|---|
| Idea → ticket → Plan mode → work (common resolution path) | task-resolution-lifecycle.md |
What to set up per repo (.cursor/, terminal pin, rules) |
repo-bootstrap.md |
.cursor/templates bundle map (minimal / standard / docs-heavy cp sets) |
templates/.cursor/README.md |
| How to write an issue or ticket agents can run | create-engineering-task.md |
| How to execute a task (modes, terminal vs agent, verify) | work-task-with-agents.md |
| Calm review pass (read code + tests, under research) | review-calm-read.md |
| Post-deploy / UI debugging (DevTools, cloud read-only, dashboards) | debug-prod-ui-flows.md |
| Model choice, value tier (re-check pricing), when to escalate | models-and-modes.md |
| Default: Plan first (summarize + plan; pad chat; Agent after) | plan-first-and-ui-context.md |
@ skills, browser capture, chat patterns |
skills-and-chat-patterns.md |
| File | Role |
|---|---|
task-resolution-lifecycle.md |
Idea, good-enough task, then Plan mode with max context and sharpened plan before execution. |
repo-bootstrap.md |
Per-repo .cursor/ checklist and pinned terminal; pasteables → templates/.cursor/README.md. |
create-engineering-task.md |
Ticket skeleton and agent-friendly fields. |
work-task-with-agents.md |
Agent vs terminal, chunking, verify. |
review-calm-read.md |
Review modus operandi: stop, read code and test results calmly (draft). |
debug-prod-ui-flows.md |
DevTools, cloud read-only, dashboards. |
models-and-modes.md |
Model policy; routine best value tier (e.g. Composer 2 non-Fast—re-validate when pricing changes). |
plan-first-and-ui-context.md |
Default bias: Plan before Agent; use chat context (no extra “skill” for built-in UI). |
skills-and-chat-patterns.md |
@ skills and capture patterns. |
device/work.md holds your SDE2-oriented personal work document outline: deploy runbook, debug stub, on-call stub, verify deployed results, test locally. Put team-specific URLs, service names, and escalation paths there.
This coding/ tree adds Cursor- and agent-specific playbooks (skills, models, DevTools flow) that apply across repos. When a procedure needs both, link from the ticket: e.g. “Debug: follow coding/debug-prod-ui-flows.md; paste dashboard links from device/work.md.”