agent pnl silently ignores unexpected positional args and falls back to 24h

[timbrinded]

Summary

hlgo agent pnl silently accepts an extra positional argument like 720h, ignores it, and falls back to the default --lookback-hours 24.

This is easy to miss in scripts/agent flows because the command succeeds and returns valid-looking JSON, but for the wrong time window.

Environment

• repo: timbrinded/hlgo
• local repro commit: aa673f0
• binary rebuilt from source on Linux arm64

Steps to reproduce

Assuming HL_PRIVATE_KEY is set and the account has fills/positions:

hlgo agent pnl 720h --format json

Actual outcome

The command succeeds, but the positional 720h is ignored. The output still reports the default lookback:

{
  "lookback_hours": 24,
  ...
}

In my repro, the same happened for 168h and 336h as well.

Expected outcome

One of these should happen instead:

1. Preferred: reject unexpected positional args with a validation error, e.g. unknown argument: 720h or accepts no args.
2. Alternative: intentionally support positional duration syntax and parse 720h / 168h / 336h into lookback_hours.

What should not happen is silent success with the default 24h window.

Likely root cause

cmd/agent_pnl.go only reads the --lookback-hours flag:

lookbackHours, _ := cmd.Flags().GetInt("lookback-hours")

and defines:

cmd.Flags().Int("lookback-hours", 24, ...)

But the command does not appear to set an Args validator (for example cobra.NoArgs), so extra positional input is accepted and ignored.

Suggested fix

• Add Args: cobra.NoArgs to newAgentPnlCmd() (and possibly audit other commands for the same silent-extra-arg pattern).
• Add a regression test covering:

hlgo agent pnl 720h

which should fail with a non-zero exit code.

Why this matters

For humans this is confusing; for agents/scripts it is worse, because the command returns plausible output and can lead to incorrect performance analysis or decision-making while appearing to work correctly.