[CFX-7713] fix(task): name the real template markers in the not-in-template error

[ajalon1]

Potentially a blocker before I release CLI 0.3.0!

RATIONALE

The "not in a template directory" error tells users to create a folder they are already standing in.

#791 replaced internal/task.IsTemplateDir (a bare os.Stat on .datarobot) with repo.IsTemplateDir, which is a strictly narrower predicate: it wants .datarobot/answers, or a .datarobot/cli holding something other than state.yaml. That consolidation was the right call — the old check treated the stray .datarobot/cli/state.yaml that older CLI versions scattered outside template directories as proof of a template.

But the error text did not move with it. So the users most likely to hit the stricter gate — the ones whose .datarobot folder exists and is no longer sufficient — get told:

This command requires a '.datarobot' folder to be present.

They have one. The message sends them looking for something they already did, and says nothing about the marker actually missing. It affects dr task list, dr run, and shell completion, since the gate runs first in Discovery.Discover, ahead of the PreferRootTaskfile fallback.

CHANGES

• Build the ErrNotInTemplate message from the same constants the predicate reads, so it names the real markers and points at dr template setup instead of leaving the remedy to guesswork.
• Add repo.TemplateDetectStateFileName. The state.yaml exemption was a bare literal in repo.IsTemplateDir; hardcoding it a second time in the message would have rebuilt the same drift this PR is fixing, so predicate and message now read one constant.

BEFORE / AFTER

Fixture: a directory containing a bare .datarobot/ folder.

Before (current main):

$ dr task list
You don't seem to be in a DataRobot Template directory.
This command requires a '.datarobot' folder to be present.
$ echo $?
1

After (this PR):

$ dr task list
You don't seem to be in a DataRobot Template directory.
This command requires a '.datarobot/answers' folder, or a '.datarobot/cli' folder holding something other than 'state.yaml'.
Run 'dr template setup' to create one, or switch to an existing template directory.
$ echo $?
1

Checked against three fixtures — a bare .datarobot/, a .datarobot/cli/ holding only state.yaml, and a bare .datarobot/ alongside a committed root Taskfile.yaml. Output is identical between main and this PR on all three except the message text, which is the intended blast radius: no behavior change, same exit code, same gate.

NOTE

There is a separate regression around dr run, discovered while testing here, but I've split that out and have moved it to CFX-7712.

TESTING

Two tests in internal/task/discovery_test.go pin the text to the predicate:

• TestFormatDiscoveryErrorNamesRealTemplateMarkers — the message must name
  every marker repo.IsTemplateDir checks.
• TestTemplateMarkersMatchAdvertisedRequirement — the directory shapes the
  message advertises must pass the gate, and the two it excludes (a bare
  .datarobot, and a .datarobot/cli holding only state.yaml) must fail it.

Both fail against the previous message. Verified by temporarily restoring the old
text: it trips the .datarobot/cli and state.yaml assertions, then passes once
the fix is back.

task lint clean across all three GOOS. Full task test green, no races.

NOTES

Related: #808 covers the other half of the release review that surfaced this (the
versioning guidance in releasing.md). Independent of this PR.

Worth a reviewer's judgment: naming state.yaml in user-facing copy leaks a
detail most people will not care about. I kept it because without it a user with
a .datarobot/cli folder has no way to understand why they still fail the check.
Happy to drop it if you would rather keep the message shorter.

---

Note

Low Risk
User-facing error copy and a shared filename constant only; template detection logic is unchanged.

Overview
Fixes the not-in-template error so it matches repo.IsTemplateDir instead of telling users a bare .datarobot folder is enough.

FormatDiscoveryError now names .datarobot/answers or a .datarobot/cli folder with something other than state.yaml, and points users at dr template setup. The state.yaml exemption is a shared TemplateDetectStateFileName constant so the message cannot drift from the predicate. Tests pin the copy to those markers and to the directory shapes that pass vs fail the gate. Detection behavior is unchanged.

^{Reviewed by Cursor Bugbot for commit fdcdada. Configure here.}

[datarobot-pr-review-router]

Code Ownership

Cli Maintainers

• internal/repo/detect.go
• internal/repo/paths.go
• internal/task/discovery.go
• internal/task/discovery_test.go

Review requested from the teams above. Labels will be removed automatically upon approval.

[carsongee]

I was worried that would have unintended consequences. I thought I had tracked all of them down. Good catch

[ajalon1]

/approve-smoke-tests

[ajalon1]

I think these tests are overkill, but hey! Tests are cheap these days, and we can always remove them or update them if template assumptions shift.

[github-actions]

🚀 Smoke tests triggered! Running on Linux and Windows...

[github-actions]

✅ All smoke tests passed!

✅ Linux: success
✅ Windows: success

View run details