project-os_skill

Overview

This skill is an AI Project OS that brings a reproducible delivery loop to new repositories: development standards, iteration logs, and release closure. It enables autonomous orchestration and rule-driven execution so agents can drive implementation, verification, release, smoke tests, and retrospectives with traceable evidence.

How this skill works

On first run the agent detects whether the target project has the required indices and docs (AGENTS.md, commands, docs/logs). With user confirmation it initializes templates and merges existing files rather than overwriting them. After initialization, the system enforces a Rulebook, exposes commands/skills/workflows for automated orchestration, and records logs and workflows for audit and rollback.

When to use it

• Bootstrapping a new project that needs a unified delivery and release workflow.
• Standardizing verification, smoke testing, and release rules across multiple repositories.
• Automating end-to-end delivery: from requirement to production smoke to retrospective.
• Enforcing auditability and rule-driven decisions for high-compliance teams.

Best practices

• Always confirm initialization when prompted so templates merge safely with existing files.
• Keep Rulebook changes centralized in AGENTS.md to ensure consistent enforcement and auditing.
• Update commands/commands.md whenever you add or modify skills to keep agent indices accurate.
• Run build/lint/tsc and smoke tests at the end of each development cycle as mandated.
• Use non-repository directories for smoke tests to avoid accidental writes into the repo.

Example use cases

• Install the skill and initialize a microservice repo to auto-generate AGENTS.md, commands, and logs.
• Enforce a release checklist across a monorepo so agents orchestrate migrations, deploys, and online smoke tests.
• Automate rollback and acceptance checks by using the skill's workflows and Rulebook to decide exceptions.
• Provide a team-wide template for iteration logs and post-release retrospectives tracked by the agent.