- name
SKILL.md
O arquivo da skill permanece no repositório do publicador.
Cole o comando copiado em um terminal. Ele instala este SKILL.md nesse agente.
npx skills add basecamp/hey-cli --skill verify-hey-cli -a cursor -yCole o comando copiado em um terminal. Ele instala este SKILL.md nesse agente.
npx skills add basecamp/hey-cli --skill verify-hey-cli -a claude-code -yCole o comando copiado em um terminal. Ele instala este SKILL.md nesse agente.
npx skills add basecamp/hey-cli --skill verify-hey-cli -a codex -yCole o comando copiado em um terminal. Ele instala este SKILL.md nesse agente.
npx skills add basecamp/hey-cli --skill verify-hey-cli -a grok -yEnvie a mensagem copiada em um chat do Grok Bot. O Bot busca este SKILL.md no GitHub e o adiciona às skills que todos os seus Bots compartilham.
Install this skill: https://github.com/basecamp/hey-cli/tree/HEAD/.agents/skills/verify-hey-cliDo publicador
Work from the hey-cli repository root. Read the feature index, then the relevant recipe and its actual test assertions. This is an operating guide, not another test suite or a command inventory.
For local configuration, follow settings; no server is needed.
Public contributors can run local CLI checks, unit tests and offline harness regressions without backend access. Live integration journeys use tests/smoke and require an authorized internal test environment: Haystack is closed source and is not supplied by this repository. Unavailable access is a blocker, not a provisioning task or permission to use a production account. Read the smoke README and development setup first. They require prepared local Haystack, exclusive development-account/server ownership, Chromium and the registered loopback OAuth client. Sending additionally requires confirmed local-only sender routing and SMTP capture. Do not reset/reseed an existing database, use production/beta, copy operator credentials or stop somebody else's server. A private client directory does not isolate the database.
make build
export VERIFY="$PWD/.agents/skills/verify-hey-cli/scripts/smoke"
The helper pins this build for one invocation. Rebuild after product or SDK changes. It records the source revision, tracked working-tree diff hash and binary hash; an explicit --binary is an artifact comparison, not proof it was built from the current checkout.
The smoke harness signs in through the real browser/OAuth flow. Each invocation gets fresh HOME/config/state/cache/work and finishes resource cleanup/readbacks before removing credentials. Do not use a bare hey, inherited profiles, a stored bearer token or an operator's normal authenticated session.
After taking ownership and establishing the setup above, require actual authenticated reads, not just HTTP /up:
"$VERIFY" TestAuthStatus TestBoxesList TestCalendarList
Require exit zero, every named test passed, and result.json with ok: true. This doctor covers authentication/mail/calendar discovery, not search indexing, SMTP routing, image processing or TUI rendering. Check the relevant prerequisite in the feature recipe too. Browser sign-in creates a server session/grant and can create a notification; it is not side-effect-free.
Each fresh invocation also repeats TestMain's server-reachability, browser-login and real OAuth setup checks before any test. Doctor each new live environment and after anything surprising. Diagnose a failed doctor before driving mutations. A healthy process with a wedged UI still needs a reset/relaunch. If the environment cannot supply the required state safely, record the attempted path and blocker; never replace the journey with an easier listing.
For a regression, follow the bug-reproduction procedure. First reproduce the reported behavior, then repeat that same check after the fix. Existing smoke cases are building blocks, not substitutes for a missing assertion.
"$VERIFY" TestDraftLifecycle TestReplyDraft
Name tests explicitly; do not pass a broad regex or run the entire mutation suite. Inspect fixtures and cleanup before selecting a case. Each invocation returns its evidence directory. --binary /absolute/path/to/hey compares a separately built revision; --tags verification compiles tagged cases only after their extra prerequisites are arranged. Do not select credential-output or logout tests through this evidence path.
The helper forces strict, uncached runs and verifies named tests actually ran without failed/skipped subtests. It does not decide whether those assertions prove the bug, choose a safe sender or grant account/server ownership.
For a TUI change, use the TUI recipe. A CLI command or Bubble Tea model test is not a rendering/key-interaction check. Unattended PTY driving is not established here; report that limitation rather than claiming CLI success covers the TUI.
Keep each run's private directory outside the checkout. It contains the binary hash and source metadata, Go JSON events, readable output, original exit status, audited result and allowlisted data-command receipts. Never record credentials, OAuth URLs/state or secret-valued arguments; do not upload private mail evidence to a PR without review.
Record the feature, reported condition, exact public entry point, fixture identity, action, expected outcome and observed outcome. A mutation needs a second read of the stored value and the feature's independent file/browser/delivery observation. An acknowledgement alone is not proof. Verify what --dry-run or --draft actually skips; a draft is a real server write.
Report passed / failed / blocked / not attempted per acceptance case, separately for CLI, model and live TUI. A skipped/missing test, wrong surface or inconclusive result is not a pass. A passing map/reference check is not behavior coverage. Record baseline and candidate separately, including setup differences; do not claim a comparison between unlike scenarios.
Cleanup is part of the result, including after expected failures. Checked smoke callbacks cover the established fixture paths, not every legacy test. They retain hidden contacts/aliases and trash mail rather than permanently deleting it. They do not revoke all OAuth grants or remove sign-in notifications; do not mass-delete somebody else's sessions/grants.
The helper removes its owned temporary files after its process group exits, including Chrome residue, unless interrupted smoke client state needs recovery; its exact path is retained in result.json. A killed suite can bypass resource cleanup: inspect its exact fixture receipts and recover those resources with an isolated development login before calling teardown proven. Do not glob-delete other runs' state. Retain failure evidence and use a fresh run after fixing the environment or helper.
Stop a backend/capture only if this run started it, by its dedicated terminal or recorded container ID. Inspect that its ports/processes stopped. Never kill by process name. Confirm evidence still exists after cleanup; /tmp survives cleanup but not reboot/tmp cleaners. Retain reviewed evidence elsewhere when needed.
# From the repository root; POSIX/Linux/macOS, Python 3, git, ps and cached Go dependencies.
.agents/skills/verify-hey-cli/scripts/smoke TestReplyDraft
.agents/skills/verify-hey-cli/scripts/test-smoke
The helper runs the existing Go harness; it is not a second HTTP/auth client. Local settings use the public binary directly as documented in their recipe.
To maintain this guide, use the index's maintenance contract. Keep recipes tied to observable behavior, not every implementation symbol. A maintenance run edits only this skill's directory. A needed fix in tests/smoke or product code is a separate change, not permission to copy that code here or weaken the expected behavior.
Reproduce and verify HEY CLI or TUI behavior against local development Haystack. Use before and after a bug fix, for PR runtime proof, or to maintain the user-facing feature map. Uses the existing smo
---name: verify-hey-clidescription: Reproduce and verify HEY CLI or TUI behavior against local development Haystack. Use before and after a bug…---# .agents/skills/verify-hey-cli/SKILL.mdVisão geral
Reproduce and verify HEY CLI or TUI behavior against local development Haystack. Use before and after a bug fix, for PR runtime proof, or to maintain the user-facing feature map. Uses the existing smoke harness, owned fixtures and private evidence; unit tests alone are not live proof.
Perguntas frequentes
Baixe .agents/skills/verify-hey-cli/SKILL.md do publicador ou copie o caminho e abra o repositório. O PlugMCP lista a skill e não hospeda o arquivo.
O publicador o mantém em https://github.com/basecamp/hey-cli.
Claude, Cursor, Codex e Grok Build conseguem ler um SKILL.md que você adiciona a partir do repositório do publicador. O Grok Bot o instala quando você envia em um chat o link da pasta da skill no GitHub.