novo-projeto

Installs an AI agent governance framework into new or existing repositories with profile selection and non-destructive updates.

1|Updated Jun 2, 2026
One-click install
npx skills add https://github.com/psiagoleal/ai-coding-agent-profiles --skill novo-projeto-psiagoleal
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: novo-projeto
Source: https://github.com/psiagoleal/ai-coding-agent-profiles/tree/main/skills/novo-projeto
Command: npx skills add https://github.com/psiagoleal/ai-coding-agent-profiles --skill novo-projeto-psiagoleal

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Adopting a shared AI agent configuration framework in a repository is risky: choosing the wrong confidentiality profile, losing pre-existing CLAUDE.md content, or having project-specific knowledge silently overwritten by future updates. This Skill guides the full adoption workflow so nothing is lost and the project stays updatable. ## Core Features & Use Cases - Profile selection via decision tree: Chooses between empresa, externo-confidencial, and pessoal profiles by asking the user, never guessing, since the choice carries confidentiality consequences. - Safe installation workflow: Enforces a git baseline commit, dry-run-first execution of setup-profile.sh, and preservation of any pre-existing CLAUDE.md content before installing. - Update-safe content placement: Directs project knowledge into USER marker islands in AGENTS.md or into docs/ files referenced from those islands, so non-destructive --update runs never erase it. - Use Case: You join a client project under NDA that already has a hand-written CLAUDE.md. The Skill walks you through picking the externo-confidencial profile, backing up and redistributing the old content, installing with --dry-run, and filling CURRENT-STATE.md so the next session starts informed. ## Quick Start Ask the agent to adopt the agent governance framework in this repository, choosing the right profile with me and preserving my existing configuration files.

Frequently Asked Questions about novo-projeto

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I install an AI agent configuration framework into an existing repository?

Create a git baseline commit first, then run setup-profile.sh with --dry-run to review the plan before applying it. If a CLAUDE.md already exists, back up its content and remove it before installing so it does not compete with AGENTS.md as the single source of truth.

How do I choose between empresa, externo-confidencial, and pessoal profiles?

Use the decision tree: company projects use empresa, client or NDA work uses externo-confidencial, and public open-source uses pessoal. When none clearly applies, default to externo-confidencial as the conservative choice, and always confirm with the user rather than guessing.

How do I keep project-specific rules from being overwritten by framework updates?

Place short commands and paths inside the USER marker islands in AGENTS.md, such as comandos-exatos and estrutura-diretorios. Put longer knowledge in separate docs/ files referenced from inside an island, since everything outside the islands is regenerated on --update.

Does setup-profile.sh work on Windows checkouts?

Yes, but use --skills-mode copy because symlinks are not portable on Windows. The default symlink mode works on Unix-like systems, and --skills-mode none installs only the neutral skills folder without an agent adapter.

What happens if someone edited framework text outside the marker islands?

The installer stores a SHA-256 baseline in .agent-profile/baseline.sha256. On --update, local edits outside islands are not overwritten; the original is kept, a .new file is written, and the conflicts are listed for manual resolution.

Why does the skill require a git commit before installing the framework?

The baseline commit lets you review every later change with git diff, distinguishing what the installer wrote from pre-existing content. Without it, non-destructive --update reviews have no reference point.