Rescued from the desktop copy of the project (now being trashed):
- docs/test-cases-zh/: Chinese e2e test case designs across 6 modules
(auth, search, dashboard/settings, namespace/governance, public pages,
publish flow) plus master case sheets in md+csv.
- docs/local-run-no-docker.md: Guide for running the full stack locally
without Docker, matching the current dev environment setup.
- Delete namespace-members-data.spec.ts (TC_MEMBER_001 in
namespace-members-management.spec.ts is equivalent and the new suite
also covers add-member dialog and batch import).
- Delete namespace-reviews-data.spec.ts (TC_NS_REVIEW_001 in
namespace-reviews-management.spec.ts is equivalent and the new suite
also covers status tabs, tab switching, and sort order).
- Trim settings-pages.spec.ts to only the reset-password navigation test;
profile/security/notification page coverage now lives in the dedicated
*-settings.spec.ts files with deeper interaction tests.
Net change: 5 fewer test cases, ~1.5 min saved per CI run.
UI changed column header style from "Skill" to "SKILL" (likely via CSS
uppercase or source string). Hard-coded 'Skill' no longer matches the
accessible name. Use /^Skill$/i etc to be resilient to either case.
Affected: TC_REVIEW_003 (pending) and TC_REVIEW_004 (approved).
getByText('Display Name') is a substring match by default, so it also
matched the description paragraph "Manage your display name and
personal information." (strict mode violation: resolved to 2 elements).
Use { exact: true } to match the field label only.
- admin-audit-log TC_AUDIT_003: replace fragile getByText regex with
table/empty-state locator (strict mode .first() was matching invisible
placeholder text in filter inputs).
- profile-settings TC_PROFILE_001: drop removed "User ID" field, assert
on Display Name + Email which are the current fields.
Migrated from local desktop work (rescued from un-pushed commit 3b970865).
64 test cases with stable semantic selectors, covering pages that were
previously untested.
Three docs that answer "where should my new test go?" for CLI-related e2e:
- cli/test/e2e/README.md: decision tree across 5 test tracks, Track B
boundary, env var setup (SKILLHUB_E2E_*), new-test checklist, CI status
- web/e2e/auto-generated/README.md: marks directory as pipeline output,
naming convention, boundary with hand-written specs
- e2e-automation/README.md: output location convention, module subdirectory
layout, boundary clarification (pipeline only produces browser-subject specs)
- Add integration tests for doctor, install, list, publish, remove, whoami commands
- Expand fake-registry with resolve/delete state capture for real assertions
- Fix update command: use correct package name @astron-team/skillhub from constants
- Refactor runUpdateCommand to accept string[] instead of fragile string splitting
- Add dependency injection to updateCommand for testable unit tests without global mocks
- Replace package.json import with codegen (scripts/generate-pkg-info.ts) to avoid
leaking devDependencies into the build artifact
- Fix startNetworkFailureServer TOCTOU race by keeping listener alive
- Add TODO markers for known help command bugs (--json not forwarded, unknown topic crash)
- Extend update integration test timeout for real npm registry checks
Ensures npm publish uses the correct registry (registry.npmjs.org) even when
global npm config points to a mirror registry (e.g., registry.npmmirror.com).
- Add publish script with env validation, git checks, and build/test/pack preflights
- Add comprehensive test suite for publish workflow (302 lines)
- Update cli/package.json with @astron-team scope and full npm metadata
- Add README.md with user-focused documentation and registry info
- Add .env.example template for publish configuration
- Add Apache 2.0 LICENSE
- Add Makefile targets for build/test/lint/typecheck/publish workflows
- All publish targets include .env.local validation
- Update installation instructions across all documentation to use @astron-team/skillhub
The issue was that new URL().pathname returns Unix-style paths on Windows
(e.g., /D:/a/skillhub/...) which Bun.spawn() cannot resolve correctly.
Using fileURLToPath() converts the URL to a proper platform-specific path.
Changes:
- cli/test/helpers/run-cli.ts: use fileURLToPath for cwd
- cli/test/integration/version-command.test.ts: same fix for build spawn
Bun.which('bun') on Windows CI returns a non-existent path
(C:\Users\runneradmin\.bun\bin\bun.exe), causing all integration
tests to fail. Add existsSync() check to fallback to process.execPath
when the resolved path doesn't exist.
Also fix version-command.test.ts to use process.execPath instead of
hardcoded 'bun' string for cross-platform compatibility.
Fixes 19 failing integration tests on Windows platform.
- Use Bun.which() with process.execPath fallback in run-cli helper to
resolve bun executable on Windows (fixes 22 integration tests)
- Normalize paths in credentials-store test for Windows backslash compat
- Use shell: true on Windows in updater spawn for proper exe resolution
Replace 'any' type annotation with proper type guard in inventory-store.ts
and add optional chaining in test to satisfy TypeScript strict checks.
All three platforms (Ubuntu, macOS, Windows) CI checks now pass.
Add 36 test cases covering:
- Path traversal and symlink attack prevention in archive extraction
- SkillHubClient error handling (401/403/404/network) for all endpoints
- Inventory store concurrent writes and stale lock recovery
- Config store read/write round-trip
- Platform utilities (package-manager, updater, paths)
- Output formatting (printResult, humanize)
Tests use cross-platform commands (node) instead of shell builtins
for CI compatibility across macOS/Linux/Windows.
Protect doctor metadata scanning from symlinked agent, skill, and .skillhub directories, and make inventory mutations use locked atomic writes with stale lock recovery.
Document the full CLI workflow so users can understand configuration precedence, install targets, local state files, troubleshooting, and local verification steps.
Enable API token authentication for CLI endpoints by adding /api/cli/
to the filter's path whitelist. Previously, CLI endpoints were not
processed by the token authentication filter, causing all Bearer token
requests to fail with 401.
Root cause: two issues caused CI failure:
1. publishSkill hit 429 rate limit from other tests in the suite
2. glob pattern '?' was ambiguous for matching literal '?' in URLs
Fix: remove publishSkill dependency, use regex route matching, and
return fully mocked API responses with fake skill data. This tests
the frontend pagination UI behavior without needing real published
skills.
The namespace detail page hardcoded a limit of 20 skills with no
pagination controls, so namespaces with more than 20 skills appeared
truncated. Add page state, wire it into the search query, and render
the shared Pagination component when total exceeds PAGE_SIZE. Also
reset page to 0 when the namespace param changes.
Closes#350
Add an "Update" button to each skill card on the My Skills dashboard
page. Clicking it navigates to the Publish page with the skill's
namespace and visibility pre-selected, reducing manual steps when
re-publishing a skill package.
- Add visibility field to SkillSummaryResponse so the list API exposes
each skill's current visibility setting
- Add publish-prefill module to normalize and validate URL search params
- Wire TanStack Router validateSearch on the publish route
- Add E2E tests covering the prefill flow and invalid-param fallback
