qa-session-ops
Skill78 starsChanged 6 days agoApache-2.0View on GitHub
Run and troubleshoot `bun run beep qa` recording sessions: record / extract / report subcommands, lane choice (playwright recordVideo vs OBS obs-websocket against real Chrome), OBS provisioning (websocket enable, PipeWire portal restore-token, scene setup), and collector/witness debugging. Use when a QA capture will not start, produces no events, records a black window, or you must pick a recording lane. The loop protocol itself lives in `browser-qa-loop`; this skill is the machinery under it.
In this repository38 other files it gives its agents.
- AGENTS.mdAGENTS.md
- explorations/AGENTS.mdAGENTS.md
- adhdSkill
- atom-reactivity-specialistSkill
- browser-qa-loopSkill
- closeoutSkill
- codex-findingsSkill
- crispenSkill
- effect-first-developmentSkill
- exploreSkill
- grillingSkill
- grill-meSkill
- grill-with-docsSkill
- jsdoc-annotation-specialistSkill
- motion-evidence-reviewSkill
- onepassword-secret-refsSkill
- ontology-architectSkill
- ontology-conceptualizerSkill
- ontology-curatorSkill
- ontology-foundational-auditorSkill
- ontology-mapperSkill
- ontology-requirementsSkill
- ontology-scoutSkill
- ontology-validatorSkill
- operator-handoffSkill
- orchestrateSkill
- plasma-sensitive-entrySkill
- portlessSkill
- quality-review-fix-loopSkill
- reflectSkill
- remote-development-agentsSkill
- schema-first-developmentSkill
- shadcnSkill
- sparql-expertSkill
- teachSkill
- turborepoSkill
- unslopSkill
- yeetSkill
The file
We split this skill into the rules it gives an agent
1 rule. About 0 are common, 0 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈0
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 64
No events flowing? Check in order: (1) curl 127.0.0.1:43117/health.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
QA Session Ops
browser-qa-loop owns when and how to run the loop. This skill owns making
the machinery work: lanes, OBS, the event collector, and the witness.
Quick health check
bun run beep qa doctor — probes ffmpeg/ffprobe/exiftool presence,
obs-websocket reachability (127.0.0.1:4455), and playwright. Exit 1 on
required-missing. resources/qa-env-doctor.sh is the standalone fallback when
the CLI itself is broken.
Lane choice
See references/lane-selection.md for the decision table. Short version:
Lane A (playwright) for every loop round — deterministic, headless, CI-able.
Lane B (OBS) is deprecated as of 2026-08-01 — it cannot run unattended
(the PipeWire portal requires human consent by design) and silently records
black video when its restore token goes stale. Its replacement for native
drag, cursor, and OS-selection classes is Lane C (Xvfb + xdotool XTEST +
ffmpeg x11grab + CDP), proven as a prototype but not yet shipped as a CLI
lane; see browser-qa-loop and the recorded-qa-acceptance packet.
OBS provisioning (Lane B — deprecated, existing sessions only)
- obs-websocket server:
server_enabled: truein~/.config/obs-studio/plugin_config/obs-websocket/config.json(edit with OBS closed — OBS rewrites the file on exit) or OBS → Tools → WebSocket Server Settings. Port 4455; password lives in that file — export it asOBS_WEBSOCKET_PASSWORDfor the driver, never commit it. - Scene:
beep qa record --lane obsprovisions thebeep-qascene and window-capture input idempotently via@beep/obsensureQaScene— do not hand-build scenes. On Wayland the input ispipewire-screen-capture-source: the FIRST run pops the portal window picker (pick the Chrome window); OBS persists theRestoreTokenin the input settings and later runs reuse it. A stale token (compositor/portal update) just re-pops the picker — that is expected, not an error. - OBS need not be running —
ensureRunningspawnsobs --minimize-to-traydetached and retry-connects. First-ever OBS launch shows the onboarding wizard once (manual step).
Collector / witness debugging
- Collector runs inside
beep qa recordathttp://127.0.0.1:43117(--portto override):GET /health(event count),GET /witness.js,POST /events,POST /mark,POST /stop. A live session writes.beep/qa/current.json(pid + port);beep qa stopandbeep qa markdiscover it there. Stale handle after a crash → delete the file. - No events flowing? Check in order: (1)
curl 127.0.0.1:43117/health; (2) witness actually injected —window.__beepQadefined in the page console; (3) CORS — the collector allows the portless app origins; a rawlocalhost:<port>origin outside the allowed set is silently dropped by the browser; (4) events aretext/plainNDJSON POSTs — a proxy rewriting content-type breaks decode (rejects are counted in/health). - Black/empty Lane B video: the portal picked the wrong window, or the window
is minimized (PipeWire delivers no frames) — keep the Chrome window visible;
re-pick via
ensureQaSceneafter deleting the input's RestoreToken. - Clock-sync confidence
lowinreport.md: the beacon was not detected (heavy first-paint delay, or beacon disabled). Extraction still ran with +250 ms window padding; re-record with--beaconon and a settled page for precise windows. Seereferences/troubleshooting.mdfor the full table.
Worktree / parallel lanes
Portless names are machine-global and duplicate registration fails loudly.
Concurrent lanes use lane-suffixed names
(portless storybook-<lane>.beep sh -c '...'); never --force over another
lane's route. One recording session per checkout at a time (pid-checked
current.json).
Environment notes
- The
chrome-devtoolsMCP (slim, default-disabled in settings) attaches to the QA Chrome via--browser-urlfor perf traces and computed-style introspection — enable it for the session, keep it disabled otherwise. - Evidence reading is
motion-evidence-review; artifact provenance lives in XMP-beepQA tags (inspect withexiftool/ffprobe).
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

