feat(sdlc): primitive taxonomy — agents, git skill, env-management

[NeuralEmpowerment]

Summary

• README: Explicit primitive taxonomy (Commands / Skills / Agents / Hooks / Lib) with dual-mode philosophy and composition diagram. Agents column added to plugin table.
• agents/env-reviewer: New read-only specialist agent. disallowedTools: Write, Edit, MultiEdit enforces least-privilege — audits env config, cannot touch files.
• skills/git: One consolidated git skill covering push, merge, worktree, fetch, and PR lifecycle. Intent-driven, headless-friendly. The explicit /sdlc:git_* commands stay for human-in-loop control.
• skills/env-management: Three-component pattern (typed config class → auto-generated .env.example → fail-fast startup validation). Full Python reference implementation using pydantic-settings with idempotent .env sync and exclude= for runtime-discovered fields.

Design philosophy

Primitives serve two contexts:

• Human-in-the-loop: developer invokes /sdlc:git_push explicitly, stays in control
• Headless workspaces: Claude auto-invokes the git skill, no human present — tool scoping and hooks are the safety layer

The same primitives work in both. The difference is configuration.

Test plan

• README renders correctly — composition diagram, dual-mode philosophy, plugin table with Agents column
• agents/env-reviewer frontmatter valid (tools/disallowedTools/model fields)
• skills/git/SKILL.md covers all git operations via ACTION parameter
• skills/env-management/references/python/generate_env_example.py runs without import errors (after adapting imports)

[Copilot]

Pull request overview

This PR introduces a comprehensive primitive taxonomy for the agentic-primitives repository, establishing a clear architectural foundation for AI agent systems with dual-mode operation (human-in-the-loop and headless workspaces).

Changes:

• Adds explicit primitive taxonomy to README with composition diagram showing how Commands, Skills, Agents, Hooks, and Lib components interact
• Introduces env-reviewer agent for read-only environment configuration audits with enforced tool scoping
• Creates consolidated git skill covering push, merge, worktree, fetch, and PR lifecycle operations
• Adds comprehensive env-management skill with full Python reference implementation for auto-generated .env files

Reviewed changes

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

Show a summary per file

File	Description
README.md	Adds primitive taxonomy section explaining Commands, Skills, Agents, Hooks, and Lib components; includes composition diagram and updates plugin table with Agents column
plugins/sdlc/agents/env-reviewer/agent.md	New read-only specialist agent for auditing environment variable configuration with explicit tool constraints
plugins/sdlc/skills/git/SKILL.md	Consolidated git workflow skill covering push, merge, worktree, fetch, and PR operations with intent-driven automation
plugins/sdlc/skills/env-management/SKILL.md	Comprehensive skill documentation for environment management pattern with three-component architecture
plugins/sdlc/skills/env-management/references/python/generate_env_example.py	Full Python reference implementation using pydantic-settings for idempotent .env generation and synchronization

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

The frontmatter uses tools: but the existing agent (browser-qa-agent.md) uses allowed-tools:. For consistency across agent definitions, this should use allowed-tools: instead of tools:.

Suggested change

	tools: Read, Grep, Glob, Bash
	allowed-tools: Read, Grep, Glob, Bash

[Copilot]

The composition diagram states the env-reviewer agent has "tools: Read, Grep, Glob only" but the actual agent frontmatter specifies "Read, Grep, Glob, Bash". The README should be updated to include Bash in the list, or the agent should not include Bash if it's truly meant to be read-only. Since Bash is listed in the agent frontmatter and can execute arbitrary commands, this is inconsistent with the "read-only" design stated in the README diagram.