Skip to content

Add agent-guide: teach AI agents how to use APM#516

Merged
danielmeppiel merged 2 commits intomicrosoft:mainfrom
cabird:add-agent-guide
Mar 31, 2026
Merged

Add agent-guide: teach AI agents how to use APM#516
danielmeppiel merged 2 commits intomicrosoft:mainfrom
cabird:add-agent-guide

Conversation

@cabird
Copy link
Copy Markdown
Contributor

@cabird cabird commented Mar 31, 2026

Summary

This adds an agent-guide/ directory at the repo root with two files that teach AI agents (and humans) how to install, use, and create APM packages.

What's included

  • SKILL.md -- Lightweight activation triggers and key rules. Tells an AI agent when to reach for APM (e.g., user asks about agent setup, project has apm.yml, team needs reproducible agent config).
  • apm-expert.agent.md -- Deep reference persona covering:
    • Installation on Linux, macOS, Windows, WSL (curl, Homebrew, Scoop, pip)
    • When to use APM vs. manual setup (decision framework)
    • Full command reference (install, compile, audit, pack, marketplace, etc.)
    • Creating packages -- directory structure, all 7 primitive formats with examples
    • Org-wide patterns -- sharing standards across repos
    • Plugin authoring workflow
    • Troubleshooting common issues

Why

The existing .github/skills/ and .github/agents/ files teach how to develop APM itself (auth patterns, CLI logging, Python architecture). There was nothing teaching an AI agent how to actually use APM as a tool -- install it, create packages, manage dependencies, etc.

Placement

I placed this at the repo root (agent-guide/) so it's immediately visible to anyone cloning the repo or browsing on github.com. Happy to move it elsewhere if the maintainers prefer a different location.

Add a new agent-guide/ directory at the repo root containing two files
that teach AI agents (and humans) how to install, use, and create APM
packages:

- SKILL.md: lightweight activation triggers and key rules for when an
  AI agent should reach for APM
- apm-expert.agent.md: deep reference covering installation on all
  platforms (Linux, macOS, Windows, WSL), the full command reference,
  package authoring with all primitive formats, org-wide patterns,
  plugin development, and troubleshooting

Placed at the repo root for immediate visibility to anyone cloning
the repo or browsing on github.com.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
@cabird cabird requested a review from danielmeppiel as a code owner March 31, 2026 17:58
Copilot AI review requested due to automatic review settings March 31, 2026 17:58
@cabird
Copy link
Copy Markdown
Contributor Author

cabird commented Mar 31, 2026

@microsoft-github-policy-service agree company="Microsoft"

Copy link
Copy Markdown
Contributor

Copilot AI left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Pull request overview

Adds an agent-guide/ directory intended to teach AI agents (and humans) how to install, use, and author APM packages, complementing the existing developer-focused guidance in .github/.

Changes:

  • Added an activation skill (agent-guide/SKILL.md) describing when to use APM and key usage rules.
  • Added a long-form "APM expert" agent persona (agent-guide/apm-expert.agent.md) covering installation, workflows, CLI reference, and package authoring.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 5 comments.

File Description
agent-guide/SKILL.md New skill file describing activation triggers and rules for using APM.
agent-guide/apm-expert.agent.md New long-form agent persona / reference guide for APM usage, commands, and package authoring.

Reorganize agent-guide/ into packages/apm-guide/ -- a proper APM
subdirectory package that users can install with:

  apm install microsoft/apm/packages/apm-guide

Architecture: skill-centric with bundled resources
- SKILL.md index with activation triggers and key rules
- 8 focused resource files (~40-150 lines each) for progressive
  disclosure: installation, workflow, commands, dependencies,
  authentication, governance, package-authoring, troubleshooting
- Thin agent persona (~50 lines) linking to the skill for depth

PROSE compliance:
- Progressive Disclosure: SKILL.md loads first; resources on demand
- Reduced Scope: each resource covers one domain
- Orchestrated Composition: composable skill + agent + resources
- Explicit Hierarchy: SKILL.md -> resources -> sections

Content additions beyond original PR:
- Authentication: token precedence, per-org, GHE, ADO, Artifactory
- Governance: policy schema, enforcement, inheritance, audit
- Dependencies: all 13+ ref formats, version pinning, MCP deps
- Package authoring: all 7 primitive types with examples
- Troubleshooting: error-to-fix lookup table

Also extends doc-sync instruction (Rule 4) to keep skill resources
synchronized when CLI commands, auth, policy, or formats change.

Co-authored-by: Chris Bird <cabird@users.noreply.github.com>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
@danielmeppiel danielmeppiel merged commit d92d3ff into microsoft:main Mar 31, 2026
6 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants