[WaterrrForever]

What

The most common measure/diff errors since 0.9.4 (PostHog cli_error, CLI project) come from how agents call them, not from bad videos:

Error agents hit	Users	Now
Unknown flag: --from / --to / -o	~5	Did you mean --start? (--end, --out; extract the other way round)
Missing required argument: --roi (measure sync)	5	sync's own message: the mouth box, and measure face to find the face
Missing required positional argument: YOURS (diff)	4	diff compares two: the original and yours + the command
no file at D:video-studio작업대상… (bash dropped a Windows path's backslashes)	7+	fix line: quote the path or write it with /

measure <file> with no command (20 users) already prints the commands to put first (#6997), so it's unchanged.

Test plan

fails without the fix: reject-unknown-flags.test.ts "names the flag that means the same here…"

• Ran each call above on the dev CLI and read the output.
• reject-unknown-flags, src/measure, diff tests; typecheck; oxlint/oxfmt.

[github-actions]

Edit accuracy: accurate 267 (base branch 267), smooth 115 of those

The gate passes.
Smoothness is reported in the artifact, not gated. A case fails only if it fails 2 of 3 runs.

Quarantined, measured but not gated (8)

• seqplainkeys-none-center-r0-nested-z100 (fixed by studio: undo/redo on a keyframed plain element sometimes leaves the box 230-400px off (edit-accuracy seqplainkeys-none-*) #6989): fail / pass / pass, passes
• seqplainkeys-none-center-r0-root-z100 (fixed by studio: undo/redo on a keyframed plain element sometimes leaves the box 230-400px off (edit-accuracy seqplainkeys-none-*) #6989): pass, passes
• seqplainkeys-none-center-r30-root-z100 (fixed by studio: undo/redo on a keyframed plain element sometimes leaves the box 230-400px off (edit-accuracy seqplainkeys-none-*) #6989): not run
• seqplainkeys-none-pct-r0-root-z100 (fixed by studio: undo/redo on a keyframed plain element sometimes leaves the box 230-400px off (edit-accuracy seqplainkeys-none-*) #6989): pass, passes
• seqplainkeys-none-pct-r30-root-z100 (fixed by studio: undo/redo on a keyframed plain element sometimes leaves the box 230-400px off (edit-accuracy seqplainkeys-none-*) #6989): not run
• seqplainkeys-none-px-r0-nested-z100 (fixed by studio: undo/redo on a keyframed plain element sometimes leaves the box 230-400px off (edit-accuracy seqplainkeys-none-*) #6989): pass, passes
• seqplainkeys-none-px-r30-nested-z100 (fixed by studio: undo/redo on a keyframed plain element sometimes leaves the box 230-400px off (edit-accuracy seqplainkeys-none-*) #6989): not run
• seqplainkeys-none-px-r30-root-z100 (fixed by studio: undo/redo on a keyframed plain element sometimes leaves the box 230-400px off (edit-accuracy seqplainkeys-none-*) #6989): not run

[jrusso1020]

Reviewed at e202feef1e122c1afb8bec3ed9308cbc3444d9a1.

No Important finding. Synonym suggestions are limited to flags actually declared by the receiving command; unknown flags remain rejected rather than becoming execution aliases. Missing diff input and mouth ROI now reach actionable guidance, and raw mangled Windows paths get the quoting/slash suggestion.

Nit: the two newly manual missing-argument checks use Fix and exit 1, where the previous required-argument validation exited 2. CLI AGENTS.md distinguishes usage errors from execution failures. Keep the improved guidance while returning usage exit 2, and pin missing-YOURS/ROI output alongside the new synonym test. This remains nonblocking: both paths still fail clearly. On POSIX, diff's existing resolve-before-probe path prevents the Windows regex hint from matching; measure's raw path does receive it.

Validation: full workspace build succeeds. CLI bun run test src/utils/reject-unknown-flags.test.ts src/measure src/commands/diff.exit.test.ts → 120 passed / 16 files. Independent flag suite → 33 passed; actual CLI calls verified human/JSON missing-YOURS guidance, ROI guidance, --from suggestion and mangled path advice. No user media or reported bug-bash films rerun. The shared flag rejection and stderr/data conventions remain in place.

Need: yes · Size: ok · Reuse: existing Fix, flag-rejection and nearest-name helpers · Simpler: bounded synonym map and explicit guidance · Structure: fine
Risk: low — corrective error messages, with a nonblocking exit-code note.

Verdict: APPROVE
Reasoning: Guidance matches the supported commands and does not silently accept mistaken flags. Focused tests and actual wrong-call outputs pass.

— hf-review