agentleFS
lucasbaburlucasbabur/cutokyo/.claude/skills/cutokyo-runtime-qa/SKILL.md

cutokyo-runtime-qa

Skill19 starsChanged 39 days agoApache-2.0View on GitHub

Run Cutokyo as a real local product against fake and installed harnesses, verify desktop and CLI journeys, and collect reproducible acceptance evidence.

The file

We split this skill into the rules it gives an agent

5 rules. About 0 are common, 4 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈4

    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
  1. Distinctive:

    Use this only against the integrated revision.

    line 8
  2. Distinctive:

    Run in this order:

    line 19
  3. Distinctive:

    Read-only or dry-run probes of installed harnesses.

    line 31
  4. Distinctive:

    Use representative desktop viewports at 1280×800 and 1536×960, plus a constrained 900×700 window.

    line 73
  5. Very specific:

    Every finding has a stable key, owner (core, harness-claude, harness-codex, harness-opencode, extensions, cli-ops, desktop, or …

    line 101

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Cutokyo runtime QA

Use this only against the integrated revision. Builder reports are context, not evidence. A missing executable, package, fixture, screenshot, or credential makes the affected verdict blocked or failed; it never becomes a pass.

Establish the test environment

Record the revision, OS, architecture, Rust/Node/npm versions, bundled SQLite version, and Claude Code/Codex/OpenCode versions. Use an isolated temporary CUTOKYO_CONFIG_HOME and CUTOKYO_DATA_HOME. Never point a destructive setup, uninstall, migration, or fixture command at the operator's real home directory.

Run in this order:

  1. cutokyo-gates selftest, then cutokyo-gates all.
  2. The repository's format, clippy, tests, dependency, TypeScript, and fixture gates.
  3. The fake-harness E2E suite, including every released spool format.
  4. CLI journeys against temporary state.
  5. React/Vite browser E2E and accessibility checks.
  6. Supplemental Haiku JEV batch via the installed jev-qa skill: validate the YAML, run the exact workflow cases once in isolated sessions, and read every screenshot. ERROR, BLOCKED, missing evidence, or unread screenshots fail JEV acceptance.
  7. Actual packaged Tauri smoke/E2E using WebdriverIO's Tauri service where the platform supports it; JEV/browser output is not a substitute.
  8. Read-only or dry-run probes of installed harnesses. Never send a paid model request merely to prove capture unless the workflow explicitly authorizes it.
  9. Packaging/install dry runs from built artifacts, not source shortcuts. Snapshot tracked files before and after: a build that rewrites source fails.

Required adversarial worlds

Ingest and storage

Exercise duplicate hook delivery, hook session-ID drift, out-of-order events, unknown parser versions, a truncated spool file, month-old backlog, byte and age caps, concurrent readers with one writer, a second writer, long reads around a checkpoint, migration from every fixture DB, downgrade refusal, online backup, tampered-backup refusal before replacement, failed-restore recovery, FTS5 search, persisted-health corruption/restart, current versus lifetime quarantine, bounded health reads under large history, retention and one/all deletion, and PRAGMA integrity_check.

Setup and adapters

For each harness, run setup dry-run, setup, repeated setup, interrupted setup, and uninstall. Separately run install-never-activated, restore with missing/empty state, repeated restore, corrupt/partial state, concurrent user edits, symlink/non-regular file refusal, and independent subsystem-cleanup failure. No managed state must be a successful no-op. Compare before/after files byte-for-byte outside the managed block. Prove one canonical settings contract or patch semantics: an omitted UI field cannot reset an unrelated guard, and unknown fields are rejected where exactness is required. Prove durable recovery intent, backup creation, permissions, origin reporting, and exact native resume ID. Version fixtures by the harness version that produced them. Unknown versions must show reduced coverage instead of silently parsing as a known format.

Product journeys

Capture onboarding, no-harness, empty-history, active ingest, degraded coverage, quarantine, lock contention, dashboard, filtered search, session detail, resume, retention preview, one-session deletion, confirmed delete-all, installed hooks/skills/ plugins/MCP inventory, MCP enable/disable, plugin verify, secret-guard findings, AI-analysis preview/confirm/cancel/error/success, doctor, and bundle. Cover keyboard-only navigation, focus restoration, reduced motion, screen-reader names, loading, empty, partial, and error states.

Use representative desktop viewports at 1280×800 and 1536×960, plus a constrained 900×700 window. This is a desktop application; do not invent a phone layout, but do prove the minimum supported window does not overflow or hide actions.

Security and privacy

Pass only synthetic secrets through headers, URLs, nested JSON, tool output, multiline text, and split stream chunks. Scan spool, database projections, logs, crash files, screenshots, proxy traces, and diagnostic bundles for those sentinels. Also include benign high-entropy values to catch destructive false positives. Proxy activation must require consent, remain visibly active, fail open for instrumentation, and never fabricate context breakdown when unavailable.

Reconciliation

For fixture sessions, independently calculate tokens, costs, agents, tools, skills, quota windows, and coverage. Dashboard and CLI totals must match. Conflicting sources use declared precedence and never double count. Costs cite the pricing snapshot and effective interval; unknown price remains unknown.

Evidence

Save exact commands and exit codes as text or JSON logs. Browser/native journeys record step-by-step actions, expected/actual outcomes, viewport, revision, and screenshots. Installed-harness probes record the executable version and which operations were genuinely exercised. Package checks record artifact hashes and install location.

Every finding has a stable key, owner (core, harness-claude, harness-codex, harness-opencode, extensions, cli-ops, desktop, or integration), severity, reproduction, expected/actual result, and evidence path. Retest the exact journey after repair and then rerun any affected aggregate checks.

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.