fix: require Copilot-specific markers for .github/ auto-detection by Giggitycountless · Pull Request #1069 · microsoft/apm

Giggitycountless · 2026-04-30T13:15:07Z

Description

Previously, the mere existence of a .github/ directory was used to infer the project uses GitHub Copilot. This is too broad — nearly every repo has .github/ for CI workflows, CODEOWNERS, issue templates, etc. A Cursor user with standard CI would get skills installed to .github/skills/ instead of .cursor/.

Fix

.github/ is now only detected as a Copilot target when it contains Copilot-specific markers:

.github/copilot-instructions.md
.github/skills/
.github/agents/
.github/prompts/

A bare .github/ with only workflows/CODEOWNERS is no longer treated as a Copilot signal.

Changes

src/apm_cli/core/target_detection.py — Replace .github existence check with Copilot marker detection
tests/unit/core/test_target_detection.py — Update existing tests to use Copilot markers, add new test verifying bare .github/ falls through to minimal

Fixes #805

How Has This Been Tested?

All 91 existing tests pass
New test test_auto_detect_bare_github_no_copilot_markers verifies bare .github/workflows/ does NOT trigger Copilot detection

Checklist

I have read the Contributing Guide
My changes follow the established code style
I have added tests that prove my fix is effective

Previously, the mere existence of a .github/ directory was used to infer the project uses GitHub Copilot. This is too broad — nearly every repo has .github/ for CI workflows, CODEOWNERS, issue templates, etc. A Cursor user with standard CI would get skills installed to .github/skills/ instead of .cursor/. Now .github/ is only detected as a Copilot target when it contains Copilot-specific markers: - .github/copilot-instructions.md - .github/skills/ - .github/agents/ - .github/prompts/ Fixes microsoft#805

Copilot

Pull request overview

Tightens APM's target auto-detection so a repo is only treated as a GitHub Copilot (VS Code) integration target when .github/ contains Copilot-specific markers, avoiding false positives for repos that only use .github/ for CI/workflows.

Changes:

Updated detect_target() to require specific Copilot markers under .github/ rather than treating .github/ existence alone as a Copilot signal.
Updated unit tests to create Copilot marker folders for .github/ detection and added a new test ensuring a bare .github/workflows/ falls through to minimal.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 3 comments.

File	Description
src/apm_cli/core/target_detection.py	Replaces `.github` existence check with a Copilot marker-based check for VS Code/Copilot auto-detection.
tests/unit/core/test_target_detection.py	Updates existing detection tests to use markers and adds coverage for the “bare .github” case.

Copilot · 2026-04-30T13:19:28Z

+    # For .github/, require Copilot-specific markers (not just CI workflows).
+    # A bare .github/ with only workflows/CODEOWNERS/etc. is NOT a Copilot signal.
+    github_copilot_markers = [
+        ".github/copilot-instructions.md",
+        ".github/skills",
+        ".github/agents",
+        ".github/prompts",
+    ]
+    github_exists = any(
+        (project_root / marker).exists() for marker in github_copilot_markers
+    )


detect_target() now treats .github/ as present only if one of the listed markers exists, but the marker set looks incomplete relative to the repo's own Copilot/VS Code integration surface. In particular, APM supports .github/instructions/ (and also .github/hooks/, plus legacy .github/chatmodes/) as first-class Copilot integration directories, so projects that only have modular instructions (or hooks) will now fall through to minimal and skip Copilot integration.

Consider expanding the marker list to include the other Copilot-specific subdirectories APM reads/writes (at least .github/instructions, and likely .github/hooks / .github/chatmodes) so auto-detection doesn't regress for those setups.

Thanks for catching this! ✅ Fixed: added , , and to the marker list. APM reads/writes to all of these directories for Copilot integration, so they should all trigger auto-detection.

Copilot · 2026-04-30T13:19:28Z

+    # For .github/, require Copilot-specific markers (not just CI workflows).
+    # A bare .github/ with only workflows/CODEOWNERS/etc. is NOT a Copilot signal.
+    github_copilot_markers = [


This change updates the semantics of VS Code/Copilot auto-detection (a bare .github/ should no longer trigger Copilot), but several docs pages still state that detection is based on .github/ directory existence alone (e.g. docs/src/content/docs/introduction/how-it-works.md, docs/src/content/docs/reference/cli-commands.md auto-detection table, and docs/src/content/docs/integrations/ide-tool-integration.md). Please update the docs to reflect the new "Copilot-specific markers" requirement, otherwise the documented behavior will be wrong for users.

@copilot apply changes based on this feedback

Great catch! ✅ Updated all three docs pages:

how-it-works.md: Changed 'falling back to ' to 'falling back to minimal (AGENTS.md only)' and added note about marker-based detection

cli-commands.md: Updated VSCode integration description to mention Copilot-specific markers

ide-tool-integration.md: Updated auto-detection callout to explain marker requirement and that bare with only CI workflows won't trigger detection

Copilot · 2026-04-30T13:19:29Z

+
+        assert target == "minimal"
+        assert reason == REASON_NO_TARGET_FOLDER
        assert "no target folder found" in reason


In test_auto_detect_bare_github_no_copilot_markers, reason is asserted equal to REASON_NO_TARGET_FOLDER and then also asserted to contain the same literal substring. The second assertion is redundant once equality is checked, and it makes future reason-message refactors noisier. Consider keeping only the constant-equality assertion (or only the substring assertion, but prefer the exported constant).

Suggested change

assert "no target folder found" in reason

Good point! ✅ Removed the redundant substring assertion. The constant-equality check is sufficient and cleaner.

- Add .github/instructions, .github/hooks, .github/chatmodes to Copilot-specific marker list per Copilot review feedback - Update 3 docs pages to reflect marker-based .github/ detection: - how-it-works.md: clarify fallback to minimal, not .github/ - cli-commands.md: update VSCode integration detection description - ide-tool-integration.md: document marker requirement explicitly - Remove redundant substring assertion in bare-github test - Add 3 new tests for instructions/hooks/chatmodes markers Addresses Copilot review comments on microsoft#1069

- tests/unit/core/test_target_detection.py: rename unused 'reason' to '_reason' in test_auto_detect_neither_folder (RUF059). - src/apm_cli/core/target_detection.py: collapse multi-line generator expression to satisfy 'ruff format'. Verification: uv run --extra dev ruff check src/ tests/ -> All checks passed! uv run --extra dev ruff format --check src/ tests/ -> 623 files already formatted uv run --extra dev pytest tests/unit/core/test_target_detection.py -> 101 passed Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

danielmeppiel · 2026-05-02T14:58:17Z

DevX UX Analysis -- detection signal design

Pushed lint fix (7ad1d44) -- RUF059 unused reason + ruff format collapse. CI should be green now.

Beyond the lint, raised the maintainer's parity question with our devx-ux-expert lens. Summary of the analysis (full critique kept locally; verdict + concrete asks below).

Verdict on the marker-list approach

Ship with adjustments. The PR fixes the reported bug (#805) and is directionally correct. Two real UX issues to address before merge.

Issue 1 -- the marker list is over-broad

Six of the seven markers (skills, agents, prompts, instructions, hooks, chatmodes) are directories APM ITSELF creates. They only exist after first apm install --target copilot. So the auto-detect mostly serves the steady-state case; first-time Copilot users still need the explicit flag.

The clean parity story: .github/copilot-instructions.md is the single Copilot-native pre-APM marker (GitHub's own docs tell users to create it; many Copilot repos already have it). That gives true parity:

Tool	Detection signal
Claude	`.claude/` (tool's namespace)
Cursor	`.cursor/` (tool's namespace)
OpenCode	`.opencode/` (tool's namespace)
Copilot	`.github/copilot-instructions.md` (Copilot's own canonical file)

Each is a marker that the TOOL itself owns, not one APM creates.

Concrete proposal: trim the marker list to .github/copilot-instructions.md only. Keep the other directories as functional install targets, but they're not detection signals -- they're APM artifacts. (If we want a belt-and-suspenders signal, prompts/ and agents/ are also Copilot-native; skills/, instructions/, hooks/, chatmodes/ are not, drop those from detect.)

Issue 2 -- the fallback hint is now wrong

After this PR, commands/compile/cli.py:501-505 and target_detection.py:253 still say:

Create .github/, .claude/, .codex/, .opencode/ or .cursor/ folder for full integration

But after this PR, creating a bare .github/ will not trigger detection. Following that hint is a trap.

Concrete fix: when minimal is selected AND .github/ exists (bare, no markers), emit a targeted hint:

[i] .github/ detected but no Copilot markers found.
    Use --target copilot to enable Copilot integration.

Mirrors the gh CLI pattern: when there's ambiguity, name it and give the user a one-copy-paste action. Don't silently degrade; narrate the decision.

Generic hint should also change from "create .github/" to either "Create .github/copilot-instructions.md for Copilot, ..." or "Use --target copilot to enable Copilot integration".

What's good

Test coverage is solid; new test_auto_detect_bare_github_no_copilot_markers is exactly the regression we want.
Docs updated in the same PR (ide-tool-integration.md, how-it-works.md, cli-commands.md) -- thorough.
.gitignore-style narrowing is the right INSTINCT vs. blanket detection.

Asks before merge

Trim the marker list (keep .github/copilot-instructions.md; optionally prompts, agents; drop the rest from detection). The other directories remain valid install targets, just not detection signals.
Update the in-code hint strings (commands/compile/cli.py:501-505 + target_detection.py:253) so they match the new detection behavior.
Add the targeted "bare .github/ -> use --target copilot" hint when minimal is chosen and .github/ exists. One log line, no new flag.

Happy to fold (1)-(3) into a follow-up commit on this branch if you'd prefer not to bounce it back -- LMK.

Copilot AI review requested due to automatic review settings April 30, 2026 13:15

Giggitycountless requested review from danielmeppiel and sergio-sisternes-epam as code owners April 30, 2026 13:15

Copilot started reviewing on behalf of Giggitycountless April 30, 2026 13:15 View session

Copilot AI reviewed Apr 30, 2026

View reviewed changes

github-actions Bot mentioned this pull request Apr 30, 2026

Stop rendering CLAUDE.md dependencies that don't exist #1048

Merged

7 tasks

This was referenced May 2, 2026

[Test Improver] test: add error-handling and display-variation tests for view command #1100

Draft

[Test Improver] Monthly Activity 2026-05 #1101

Open

danielmeppiel and others added 2 commits May 2, 2026 16:53

Merge branch 'main' into fix/github-copilot-detection

81b3ce9

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

fix: require Copilot-specific markers for .github/ auto-detection#1069

fix: require Copilot-specific markers for .github/ auto-detection#1069
Giggitycountless wants to merge 4 commits intomicrosoft:mainfrom
Giggitycountless:fix/github-copilot-detection

Giggitycountless commented Apr 30, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

Copilot AI Apr 30, 2026

Uh oh!

Giggitycountless Apr 30, 2026

Uh oh!

Copilot AI Apr 30, 2026

Uh oh!

Giggitycountless Apr 30, 2026

Uh oh!

Giggitycountless Apr 30, 2026

Uh oh!

Copilot AI Apr 30, 2026

Uh oh!

Giggitycountless Apr 30, 2026

Uh oh!

danielmeppiel commented May 2, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Conversation

Giggitycountless commented Apr 30, 2026

Description

Fix

Changes

How Has This Been Tested?

Checklist

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Copilot AI Apr 30, 2026

Choose a reason for hiding this comment

Uh oh!

Giggitycountless Apr 30, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 30, 2026

Choose a reason for hiding this comment

Uh oh!

Giggitycountless Apr 30, 2026

Choose a reason for hiding this comment

Uh oh!

Giggitycountless Apr 30, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 30, 2026

Choose a reason for hiding this comment

Uh oh!

Giggitycountless Apr 30, 2026

Choose a reason for hiding this comment

Uh oh!

danielmeppiel commented May 2, 2026

DevX UX Analysis -- detection signal design

Verdict on the marker-list approach

Issue 1 -- the marker list is over-broad

Issue 2 -- the fallback hint is now wrong

What's good

Asks before merge

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants