Pick a category, follow the steps, ship the work.
Turn a one-line topic into a structured slide deck. Outline the narrative, generate per-slide copy, and export to PPTX or HTML.
Check: Deck opens cleanly in target tool
Go from an empty directory to a building, tested project skeleton with a single prompt and your conventions.
Check: Build + tests pass on first run
Land in a new repo and get a working mental model fast. Map architecture, find entry points, produce a navigation guide.
Check: CODEBASE_MAP committed and shared
Turn your source modules into maintained API reference docs, wired straight into your docs pipeline.
Check: Generated docs match source behavior
Let OpenCode drive a real browser to test, scrape, or verify web flows. Connect the Playwright MCP server and capture regression artifacts.
Check: Snapshot and screenshot captured
Connect a Playwright or database MCP server so the agent can drive a browser or query live data on its own.
Check: Tool calls visible in the log
Capture your team's coding conventions in an AGENTS.md file so every agent run on the repo follows the same rules.
Check: New agent sessions follow the rules
Stop writing forgettable commit messages. Generate Conventional Commits, group staged changes into logical commits, keep history reviewable.
Check: Each commit builds and tests independently
Safely rename a core service across 45+ files. Use agents, the import graph, and git checkpoints to refactor without breaking changes.
Check: All renames complete; tests pass
Trace a subtle race condition from a stack trace to the root cause, then apply a verified fix that survives load.
Check: Tests pass under repeated load
SSH into a misbehaving server and get a diagnosis you can act on. Read logs, correlate with infra changes, propose a fix without disruption.
Check: Metrics return to baseline; post-incident written
Package a repeatable multi-step workflow as an OpenCode Skill the agent runs on demand, every time.
Check: Skill triggers automatically on cue