smoke-testing-snippets
SkillDev toolsUse when smoke-testing opencode-snippets behavior, especially PTY-driven OpenCode sessions, snippet UX, and end-to-end snippet features.
Available today. Use it from your connected AI after setup.
No other account needed.
Connect ahel once, and every AI you use reads what you have installed.
Then ask your AI: use the smoke-testing-snippets skill
What this skill tells your AI
The instructions your AI receives, as published by josxa/opencode-snippets in .opencode/skill/smoke-testing-snippets/SKILL.md and read by ahel’s review.
Use this skill before spawning opencode to verify snippet behavior end to end.
Ambient Rules
These are always relevant. Do not hide them behind progressive disclosure.
- Rebuild first with
bun run build. Localfile:///.../opencode-snippetsTUI loads builtdistoutput, so stale builds give fake negatives. - For non-interactive sanity checks, prefer
opencode run "..."before burning time on full TUI sessions. - For real TUI checks, use
pty_spawnpluspty_snapshotorpty_snapshot_wait, notpty_read. - Wait for a stable anchor like
autoeditbefore sending prompt input. - Be patient with keypresses. Use separate
pty_writesteps with tiny waits when verifying normal UX. One-burst#query + arrow + enter/tabwrites are harsher than human typing and should be treated as a separate stress case. - Treat PTY OpenCode sessions as disposable on Windows. Bun segfaults and exit code
3are host noise unless a calmer repro shows the same bug. - PTY is bad at mouse-hover validation. If a bug only appears in desktop interaction, suspect hover-state or synthetic mouse events.
- Smoke test
#skill(...)both directly and through snippets that expand into#skill(...).
When debugging PTY-driven OpenCode sessions more deeply, please read PTY + TUI learnings.
For plugin internals, load the opencode-plugin-dev skill via the skill tool.
Fast Path
- Spawn OpenCode and wait for
autoedit. - Verify normal behavior with paced
pty_writesteps first. - Then try harsher burst writes to separate real UX bugs from PTY ordering artifacts.
- If counts look wrong, inspect matcher output directly before blaming rendering.
- For non-TUI snippet features like
#skill(...), inspect stored session data too. Visible placeholder text and part metadata can prove transform happened, but they do not alone prove model-facing injected context worked.
What This Skill Helps With
- snippet filtering mismatches between expected and actual hits
- keyboard navigation vs mouse-hover fights in filtered dropdowns
- PTY/TUI verification flow for
#,#o,#oc, arrows, tab, enter - separating matcher bugs from render or scroll-state bugs
- regular chat and
opencode runsmoke tests for snippet features like#skill(...) - checking stored session/export data when visible transcript differs from model behavior
Signals
- GitHub stars
- 81
- Forks
- 8
- Last commit
- Sep 2026
Advanced
- Catalog kind
- skill
- Gateway key
smoke-testing-snippets- Source
- github.com/josxa/opencode-snippets