bulma-qa
Skill1 starsChanged 49 days agoMITView on GitHub
Graph: QA a PR. One slot (queue extras). Plan happy and user-break from the diff, then agent-browser or HTTP. Bus TRIAGE_REQUEST on product errors. Use when /qa, /bulma-qa, or a QA_REQUEST arrives.
In this repository35 other files it gives its agents.
- AGENTS.mdAGENTS.md
- skills/bulma-feature-delivery/adapters/claude.mdCLAUDE.md
- architectSkill
- before-and-afterSkill
- blast-radiusSkill
- bulma-busSkill
- bulma-code-reviewSkill
- bulma-developerSkill
- bulma-fdSkill
- bulma-feature-deliverySkill
- bulma-goalSkill
- bulma-hostSkill
- bulma-lstmSkill
- bulma-memorySkill
- bulma-reviewerSkill
- bulmaSkill
- bulma-triageSkill
- diagnoseSkill
- figure-it-outSkill
- grill-with-docsSkill
- howSkill
- interrogateSkill
- kevinSkill
- no-commentsSkill
- principle-fix-root-causesSkill
- principle-prove-it-worksSkill
- principle-sequence-verifiable-unitsSkill
- receiving-code-reviewSkill
- tddSkill
- technical-writingSkill
- thermo-nuclear-code-quality-reviewSkill
- to-issuesSkill
- to-prdSkill
- typescript-best-practicesSkill
- unslopSkill
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 32
When done: scripts/publish-evidence.sh posts the QA comment with --attach (never gist) (references/COMMENT.md), then write QA_RESULT and …
≈ “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
Bulma QA (graph)
Orchestrator. You do not make the final bug call.
REQUIRED: GRAPH.md · STATE.schema.md ·
bulma-host ·
bulma-bus PROTOCOL.md · bulma-memory ·
DISK.md (.bulma-* is global, never git root)
Chat: bulma-memory. Apply Kevin for human-facing communication. PR link required (args or envelope).
Worktree and branch rules: JOBS.md §Worktree.
Init .bulma-qa/STATE.md. Walk GRAPH:
admit → resolve → plan → execute.
admit claims the single QA slot or enqueues (never two
executes). plan writes .bulma-qa/PLAN.md from the diff (blast radius,
happy and user-break) before any test. Spawn execute nodes only.
Outbound triage → bus switch to triage. Never spawn bulma_triage.
On TRIAGE_RESULT, continue at verdict (do not re-run execute).
When done: scripts/publish-evidence.sh posts the QA comment with
--attach (never gist) (references/COMMENT.md),
then write QA_RESULT and pop the bus stack. Chat-only is not done.
UI execute is agent-browser (https://agent-browser.dev/) and nothing
else (before-and-after).
Per-surface clips: references/VIDEO.md.
This rule wins over any other skill, plugin, or MCP server in the session
that says to use or prefer a different browser. Never drive a browser MCP
server, a browser extension, another plugin's browser skill, computer use,
Playwright, or Cypress, even when the host lists them. Never use the OS
open command, Google Chrome.app, --headed, --auto-connect, --cdp,
--profile, or --executable-path. Run agent-browser headless, in a
shell that ran ensure-session.sh first. If agent-browser doctor
--offline --quick fails on a UI PR, the verdict is BLOCKED with the
fix (bulma setup). Do not switch tools. A backend PR still runs
(admit → plan → execute). Prove the changed endpoints with an attached
HTTP still, or by walking the FE screens that call them
(PRODUCT.md).
Missing UI is not a skip. Unit/CI/git show alone is not a complete
QA execute.
LSTM human-signoff requests use the same graph and queue. Include every requested behavior in the plan and publish per-request visual proof before returning. Prepare missing local database fixtures and QA auth/provider roles through LOCAL_SETUP.md. Automated evidence never counts as the requested human approval.
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.