Problem
desktest logs supports --steps ranges for viewing trajectory logs, but lacks filtering and output format options that would make it more useful for quick diagnosis and tooling integration.
Proposal
Add the following flags to desktest logs:
--summary — One-line-per-step overview showing the action taken and success/fail status:
Step 1: screenshot ✓
Step 2: click(450, 320) ✓
Step 3: type("hello") ✗ (timeout)
--failures — Show only failed steps, useful for quick diagnosis of long runs
--json — Output structured JSON for piping to other tools (jq, dashboards, CI reporters)
Notes
- These are additive flags on the existing
logs subcommand
--summary and --failures can be combined
--json output should include step number, action, result, duration, and any error message
Problem
desktest logssupports--stepsranges for viewing trajectory logs, but lacks filtering and output format options that would make it more useful for quick diagnosis and tooling integration.Proposal
Add the following flags to
desktest logs:--summary— One-line-per-step overview showing the action taken and success/fail status:--failures— Show only failed steps, useful for quick diagnosis of long runs--json— Output structured JSON for piping to other tools (jq, dashboards, CI reporters)Notes
logssubcommand--summaryand--failurescan be combined--jsonoutput should include step number, action, result, duration, and any error message