NetFabric/intelligentium / .agents/skills/apm-producer
apm-producer
FlaggedAuthor and publish APM packages: create primitives, compile, pack, and distribute. Use when: apm plugin init; authoring under .apm/ (skills, prompts, agents, instructions, hooks, commands); SKILL.md authoring; .prompt.md; .agent.md; .instructions.md frontmatter; apm compile strategies (distributed vs single-file); apm preview; apm view; apm pack bundles; apm publish; apm marketplace init / add / update; plugin.json; marketplace block in apm.yml; apm lifecycle; package-relative links; versioning; releasing from CI; dev-only primitives; devDependencies. Covers: .apm/ layout, primitive-to-target compatibility matrix, compilation placement, bundle format, marketplace.json. DO NOT USE FOR: installing packages from others (use apm-consumer); org-wide policy enforcement (apm-policy.yml).
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 4.6 KB |
| references/compile.md | markdown | 3.9 KB |
| references/pack-and-publish.md | markdown | 5.3 KB |
| references/primitives.md | markdown | 4.5 KB |
references/primitives.md
View on GitHub# APM Producer — Primitives
## Instructions
Always-on rules scoped to file globs; compiled into harness context files on every agent turn.
**Source:** `.apm/instructions/<name>.instructions.md`
```markdown
---
description: Apply TypeScript coding standards
applyTo: "**/*.ts" # glob; omit for repo-wide
---
# TypeScript Standards
...
```
| Frontmatter Field | Required | Notes |
|-------------------|----------|-------|
| `description` | Yes | Shown in compiled output; helps the agent know when to apply |
| `applyTo` | No | Glob (e.g. `src/**`, `**/*.py`); omit → repo-wide |
Compilation:
- With `applyTo`: placed in the subdirectory whose path best matches the glob.
- Without `applyTo`: placed in the root compiled file.
- Target output: `.github/instructions/`, `.cursor/rules/*.mdc`, `.claude/rules/`, `.kiro/steering/`, etc.
## Prompts
Reusable, parameterized AI workflows. The same `.prompt.md` becomes a Copilot prompt and a Claude `/command`.
**Source:** `.apm/prompts/<name>.prompt.md`
```markdown
---
description: Review this code for security issues
mode: ask # ask | agent | edit
tools: # optional tool list
- codebase
---
Review the following code for OWASP Top 10 vulnerabilities:
${selection}
```
| Frontmatter Field | Notes |
|-------------------|-------|
| `description` | Required; displayed as the command description |
| `mode` | `ask` (default), `agent`, or `edit` |
| `tools` | Whitelist of tools the prompt may use |
Per-harness output:
- **Copilot**: stays as `.prompt.md` in `.github/prompts/`
- **Claude**: compiled to `.claude/commands/<name>.md` (becomes a `/command`)
- **Cursor**: compiled to `.cursor/agents/<name>.mdc`
- **Gemini**: compiled to `.gemini/commands/<name>.toml`
## Agents
Specialized personas with their own model, system prompt, and tool boundaries.
**Source:** `.apm/agents/<name>.agent.md`
```markdown
---
name: security-reviewer
description: Expert in OWASP Top 10 and secure coding practices
model: claude-opus-4-5
tools:
- codebase
- web_search
---
You are a security-focused code reviewer...
```
| Frontmatter Field | Notes |
|-------------------|-------|
| `name` | Identifier used by harnesses |
| `description` | What the agent does; trigger for invocation |
| `model` | Optional model override |
| `tools` | Optional tool whitelist |
Supported by: Copilot (native), Claude (native), Cursor (compiled), Codex (compiled). Not supported by Gemini or OpenCode.
## Skills
Multi-file capabilities loaded on demand by harnesses that support agent skills. Built around `SKILL.md` as the entry point.
**Source:** `.apm/skills/<name>/SKILL.md` (+ sibling reference files, scripts, assets)
```markdown
---
name: my-skill # kebab-case; must match folder name
description: <trigger phrases; ≤1024 chars; include exclusions>
---
# My Skill
...
```
Skills deploy to `.agents/skills/<name>/` and are harness-neutral (`native` on all targets). Sibling files in the skill folder are deployed alongside `SKILL.md`.
## Hooks
Lifecycle handlers triggered by harness events.
**Source:** `.apm/hooks/<name>.json`
```json
{
"event": "PreToolUse",
"command": "./scripts/pre-tool-check.sh"
}
```
| Field | Notes |
|-------|-------|
| `event` | `PreToolUse`, `PostToolUse`, `Stop`, or harness-specific events |
| `command` | Script path; must be executable |
- **Claude**: hooks merged into `.claude/settings.json` under `hooks`.
- **Copilot**: hooks written to `.github/hooks/`.
- **Kiro**: individual JSON files in `.kiro/hooks/`.
- **Windsurf**: not supported.
## Plugin Type (`plugin.json`)
A package may ship a `plugin.json` at its root instead of (or alongside) `.apm/`. APM normalizes plugin format at install time into the same primitives above.
```json
{
"name": "my-plugin",
"version": "1.0.0",
"description": "...",
"skills": [...],
"prompts": [...],
"instructions": [...]
}
```
`apm pack` generates `plugin.json` from `apm.yml` + `.apm/` content automatically. Do not hand-author `plugin.json` for new packages; use the `.apm/` layout.
## Dev-Only Primitives
Primitives useful to the package author but excluded from shipped artifacts (release checklists, internal debug agents, test fixtures):
1. Author them outside `.apm/` (e.g. `dev/skills/release-checklist/SKILL.md`).
2. Declare them under `devDependencies.apm:` using a local path:
```yaml
devDependencies:
apm:
- path: ./dev/skills/release-checklist
```
3. `apm pack` excludes all `devDependencies`; `apm install --dev` deploys them locally.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |