Generate and open Inspector with one short command

Traceability

• Spec ID: inspector-zero-argument-quickstart
• Status: Implemented
• Extends Open a rendered Inspector without hunting for its output path.

Intent

The public Inspector entry point should optimize for the common developer task: generate the current repository's report and open it. Requiring callers to know the render subcommand, repeat the current workspace, and discover --open makes the first successful run unnecessarily procedural.

Running npx @qoder-ai/better-harness inspector with no Inspector arguments therefore becomes the short path for rendering and opening the current workspace. Its default evidence window covers the latest 30 UTC calendar days, with higher but still bounded commit and session caps. Explicit commands and options remain available for automation and narrower inspection.

Acceptance Scenarios

• AC-1: better-harness inspector with no Inspector arguments renders the current working directory and asks the platform opener to open the written self-contained HTML report.
• AC-2: The zero-argument path collects activity from the latest 30 UTC calendar days, bounded to at most 200 commits and 100 hydrated sessions. Date contains days with observed evidence; it does not manufacture empty-day evidence.
• AC-3: The zero-argument summary reports the resolved output path and opened: true or opened: false; an unavailable handler does not turn a successfully written report into a failed command.
• AC-4: better-harness inspector --help and the canonical Inspector help path remain read-only and print the zero-argument equivalence without collecting workspace evidence or opening a handler.
• AC-5: Explicit render and option-led invocations retain their current semantics. They open a handler only when the caller passes --open, so CI and scripts do not acquire a new side effect.
• AC-6: The bilingual Inspector guide leads with the npx zero-argument command and keeps the expanded command available for callers that need bounds or an explicit output path.
• AC-7: The public /inspector/ page presents the zero-argument npx command in the hero before the interactive sample and uses the same short command in its local-project callout.

Non-goals

• Removing render, --workspace, --open, or any existing bound.
• Opening reports for other Better Harness commands.
• Treating a browser-launch failure as a render failure.
• Changing report contents, collection bounds, evidence semantics, or the default output location.

Plan and Tasks

1. Make the Inspector CLI owner translate an empty argument list into a render of . with --open, a deterministic 30-day UTC range, and the bounded 200-commit/100-session collection limits before normal command parsing.
2. Keep help detection ahead of collection and leave every non-empty invocation on its existing parsing path.
3. Add behavior tests with an injected working directory and opener so the default can be verified without launching a real browser in CI.
4. Update English and Simplified Chinese Inspector documentation, promote the command on the public /inspector/ page, and regenerate the documentation link graph.

Test and Review Evidence

• AC-1/AC-2/AC-3/AC-5: focused Vitest coverage in test/reporting/harness-inspector.test.mjs asserts the written artifact, injected opener call, summary, and unchanged explicit-render behavior.
• AC-4: existing subprocess help coverage plus a focused assertion for the documented zero-argument default.
• AC-6: npx vitest run test/skills-docs/doc-link-graph.test.mjs and the regenerated docs/better-harness-doc-links.mmd.
• AC-7: production documentation build plus browser checks of English and Simplified Chinese /inspector/ routes at desktop and narrow widths.
• Focused runtime regression: 172 tests passed across the Inspector facade, report, session analysis, commit/session linking, and cross-provider suites.
• Documentation regression: 6 link-graph tests passed; the English and Simplified Chinese production sites built successfully, and both /inspector/ routes rendered the short command twice with the hero copy before the interactive sample, no horizontal overflow, and no console issue.
• Full regression: npm run check passed 94 files / 1324 tests, followed by package verification of 521 npm entries and 543 runtime-zip entries.

Risks

• Side effect: zero arguments now request a platform handler. The change is isolated from help and every explicit invocation, and handler failure remains best effort.
• Working-directory drift: the injected/default workspace must resolve . from the caller's working directory, not from the installed package.
• Evidence bounds: very active repositories can exceed 200 commits or 100 sessions within 30 days. The report remains explicitly bounded and must not present missing activity as observed zero.
• Cross-platform behavior: the existing openRenderedReport platform seam remains the only launcher; this change does not introduce shell-specific commands.