[BUG] the .github/ directory as a signal for auto-detecting AI agent isn't ideal

[lirantal]

Describe the bug
So I realize that you are using .github/ directory existence to decide that the AI agent is GitHub Copilot but that's a very weak signal given that .github has so many things in it like workflows, configs for CI, repo templates, etc.

Can we please change and stop that logic happening by default?

To Reproduce
Steps to reproduce the behavior:

1. Launch Cursor IDE
2. Repo has .github/ directory with workflows inside it
3. Run apm install
4. Watch skills get added in .github/skills (mistakenly because I am using Cursor, not GitHub Copilot as my AI agent)

Expected behavior
A clear and concise description of what you expected to happen.

Environment (please complete the following information):

• OS: [e.g. macOS, Linux, Windows]
• Python Version: [e.g. 3.12.0]
• APM Version: [e.g. 0.1.0]
• VSCode Version (if relevant): [e.g. 1.80.0]

Logs
If applicable, add any error logs or screenshots.

Additional context
Add any other context about the problem here.

[github-actions]

Triage decision

accept

Proposed labels

theme/portability
area/multi-target
type/bug
status/accepted
priority/high
good first issue

Milestone

0.9.4

Suggested next action

Replace the .github/ directory existence check with a more specific Copilot signal (e.g. presence of .github/copilot-instructions.md or .github/skills/) and add a detection precedence design doc comment to the issue before implementing.

Suggested issue comment

Thanks for reporting this -- the diagnosis is accurate. Using `.github/` directory existence as the Copilot detection signal is too broad: almost every serious repository has a `.github/` directory for workflows, CODEOWNERS, Dependabot config, and so on. A Cursor user with a standard CI setup gets their skills installed to the wrong target.

The fix is to replace the broad `.github/` check with a more specific Copilot signal. Good candidates:
- Presence of `.github/copilot-instructions.md` (explicit Copilot config)
- Presence of `.github/skills/` or `.github/agents/` (APM-deployed content for Copilot)

The detection precedence should also be made explicit. A reasonable order:
1. Explicit `--target` flag (always wins)
2. Presence of target-specific marker files (`.cursor/` for Cursor, `.claude/` for Claude, `.github/copilot-instructions.md` for Copilot, etc.)
3. No detection fallback -- prompt the user or require `--target`

If multiple markers are present (Cursor + Copilot both detected), APM should install to all detected targets rather than guessing one.

The detection logic lives in the target auto-detection module. This is a focused, well-scoped fix -- if you would like to take it as a PR, that would be very welcome.

Per-lens notes (collapsed)

DevX UX Expert -- User-Need Reviewer

This is a real and significant DX bug: Cursor users with a .github/ directory get skills installed to the wrong target silently. The user expectation (from any multi-target tool mental model) is that the tool detects the active environment accurately. Using .github/ as a Copilot signal is a heuristic that breaks for the majority of repositories that are not Copilot-first. The fix should use a more specific signal or require explicit --target when detection is ambiguous.

Supply Chain Security Expert -- Risk-Surface Reviewer

A misidentified target causes skills to be deployed to an unintended AI agent. The security implication is mild (skills are installed, not auto-executed), but the mis-deployment is a correctness failure that could cause confusion in regulated or enterprise environments where the target is significant. No specific supply-chain risk surface beyond the DX issue.

OSS Growth Hacker -- Contributor-Tone Reviewer

lirantal has author_association: NONE -- this is their first (or very few) interaction with microsoft/apm. The issue is well-written and technically precise, showing they understand the codebase. The response should warmly acknowledge the quality of the report, confirm the fix direction, and explicitly invite them to contribute the PR. The issue is scoped well enough to be a good first issue.

Python Architect -- Architecture Reviewer

The auto-detection heuristic change touches the target detection module, which is called at the start of every install/compile flow. The design question -- what signals are specific enough per target, and what is the precedence when multiple targets are detected -- should be answered before the PR lands to avoid follow-up regressions. The issue itself is accepted (a maintainer already labeled it accepted); the design can be resolved in a design comment on this issue rather than a separate design doc, given the contained scope.

Doc Writer -- Documentation Reviewer

The detection behavior change should be documented in the install guide (what signals APM uses to detect targets, how to override with --target). area/docs-site was considered but omitted to stay within the 6-label cap; the implementing PR should include a doc update to the install guide as part of the PR checklist.

APM CEO -- Triage Arbiter

Real bug causing mis-deployment for Cursor users with standard CI setups. Accept at priority/high for 0.9.4: this is a correctness failure in the core install flow, and the fix is well-scoped. Tagging as good first issue since lirantal has identified the exact detection location and the fix is a targeted heuristic replacement. Growth Hacker tone recommendation: explicitly invite the reporter to submit the PR.

{
  "decision": "accept",
  "decision_detail": "",
  "theme": "theme/portability",
  "areas": ["area/multi-target"],
  "type": "type/bug",
  "status": "status/accepted",
  "priority": "priority/high",
  "preserved_labels": ["good first issue"],
  "milestone": "0.9.4",
  "next_action": "Replace the .github/ directory existence check with a more specific Copilot signal and document the target detection precedence order before implementing.",
  "comment_markdown": "Thanks for reporting this -- the diagnosis is accurate. Using `.github/` directory existence as the Copilot detection signal is too broad: almost every serious repository has a `.github/` directory for workflows, CODEOWNERS, Dependabot config, and so on. A Cursor user with a standard CI setup gets their skills installed to the wrong target.\n\nThe fix is to replace the broad `.github/` check with a more specific Copilot signal. Good candidates:\n- Presence of `.github/copilot-instructions.md` (explicit Copilot config)\n- Presence of `.github/skills/` or `.github/agents/` (APM-deployed content for Copilot)\n\nThe detection precedence should also be made explicit. A reasonable order:\n1. Explicit `--target` flag (always wins)\n2. Presence of target-specific marker files (`.cursor/` for Cursor, `.claude/` for Claude, `.github/copilot-instructions.md` for Copilot, etc.)\n3. No detection fallback -- prompt the user or require `--target`\n\nIf multiple markers are present (Cursor + Copilot both detected), APM should install to all detected targets rather than guessing one.\n\nThe detection logic lives in the target auto-detection module. This is a focused, well-scoped fix -- if you would like to take it as a PR, that would be very welcome."
}

---

Triage status: agentic proposal pending human ratification.
Silence is approval. Maintainers can:

• Override any label or milestone above by editing it directly --
  human edits are authoritative and will not be reverted on
  subsequent runs.
• Re-trigger triage by applying the status/needs-triage label, or
  by removing status/triaged to enroll the issue in the next
  daily sweep.

Posted by the Triage Panel workflow. See .apm/skills/apm-triage-panel for the panel skill.

Generated by Triage Panel · ● 2.3M · ◷