ribeirogab

Scaffold or audit the memex (vault + AGENTS.md + spec templates + bundled skills) in any repo — an externalized, navigable project memory for agents (Claude Code, Codex, Cursor, OpenCode, etc.). Agent-agnostic. Idempotent — safe to run repeatedly. Use when the user wants to set up, verify, or fix the memex in a project.

Sync an installed memex with upstream — reconcile the scaffolded files against the current upstream memex: auto-apply changes to files you never touched, agent-merge the ones you edited, and never touch living vault content. Use when upstream memex changed scaffolded content (a renamed term, a reworded skill) and you want it pulled in without clobbering local edits.

Sync an installed memex with upstream — reconcile the scaffolded files against the current upstream memex: auto-apply changes to files you never touched, agent-merge the ones you edited, and never touch living vault content. Use when upstream memex changed scaffolded content (a renamed term, a reworded skill) and you want it pulled in without clobbering local edits.

You MUST use this before any creative work - creating features, building components, adding functionality, or modifying behavior. Explores user intent, requirements and design before implementation.

Use when you have an approved design (design.md) for a multi-step task — produces the fused technical spec.md + tasks.md, before touching code.

You MUST use this before any creative work - creating features, building components, adding functionality, or modifying behavior. Explores user intent, requirements and design before implementation.

Use when you have an approved design (design.md) for a multi-step task — produces the fused technical spec.md + tasks.md, before touching code.

Review a branch diff (or any diff/files pointed at) with specialized find-only subagents — project-law (rules/constitution/conventions), spec-conformance (the spec's AC-N), and documentation-consistency (stale or missing docs after the change) — merged into one plain-text verdict that reaches lgtm only when every lane is clean. Classify findings (blocker/suggestion/nitpick/question). Portable: no dependency on a native code-review tool. Trigger on 'review this branch', 'code review', 'review the diff', 'review again', or the delivery step of the spec flow.

You MUST use this before any creative work - creating features, building components, adding functionality, or modifying behavior. Explores user intent, requirements and design before implementation.

Analyze the vault for missing related: frontmatter cross-links and present suggestions interactively. Conservative: only surfaces high-evidence (wikilink already in body) and medium-evidence (filepath/title in body, or shared title/H2 terms) candidates. Never edits without explicit per-item user confirmation. Use when the user asks to find missing related: links in the vault, when /memex-sweep flags an isolated spec, or as part of the After-completing-a-spec reflection.

Search the vault for context on a topic

Use when you have a spec or requirements for a multi-step task, before touching code

Use when a user wants guidance on starting, contributing to, growing, governing, funding, securing, or sustaining an open source project, or asks about contributor onboarding, community health, maintainer burnout, code of conduct, metrics, legal basics, or open source project adoption.

Refactor or audit an existing skill against Anthropic best-practices while preserving its purpose, trigger behavior, and output format. Use when the user asks to improve, refactor, audit, or update a skill — including English phrases like "refactor this skill", "audit skill X", "apply best practices to skill", "improve this skill", "skill audit", "make this skill cleaner", AND Portuguese phrases like "melhorar skill", "atualizar skill", "auditar skill", "aplicar boas práticas", "refatorar skill", "limpar skill". This skill does NOT create new skills (use `skill-creator` for that) and does NOT change WHAT a skill does — only HOW it's organized.

Clone repos using bare clones + git worktrees, develop changes, deliver via Pull Requests. Use this skill whenever the user asks you to clone, code, fix, edit, refactor, implement a feature, or open a PR on any repository — including PT-BR phrasings like "clonar repo", "refatorar", "abrir PR", "corrigir bug", "implementar feature". Covers branch naming, quality gate discovery, commit conventions, and PR delivery rules.

Conduct a specwright milestone: read the board, dispatch every ready issue to an issue-owner sub-agent (parallel, one worktree each), track progress, apply circuit breakers, and close out with goal reconciliation, learnings promotion, and a final report. Resumable from any fresh session. Trigger on '/sw:run', 'run the milestone', 'continue the milestone', or when the user asks to resume conducting a milestone.

Open the pull request for the current issue, following specwright PR conventions — resolves branch/base, pushes if needed, fills the repo PR template, English Conventional-Commit title and body, records the runtime-verification results, no AI attribution. Use as the delivery step of the issue pipeline, or whenever the user asks to open a PR.

Set up specwright's per-repository content in the current repo: the .specwright/ vault (conventions/, issues/, milestones/), an entry point stating the sw plugin requirement, and the .gitignore lines. Always asks a commit mode: shared (committed CLAUDE.md and committed vault) or local (git-ignored CLAUDE.local.md and git-ignored vault, for a repo you do not own). Content-only — writes no machine configuration. Idempotent, safe to re-run. Trigger on '/sw:init', 'set up specwright here', 'initialize specwright in this repo'.

You MUST use this before any creative work - creating features, building components, adding functionality, or modifying behavior. Explores user intent, requirements and design before implementation, concludes whether the work is a single issue or a milestone (a large delivery decomposed into issues), and writes the issue/milestone artifacts.

Use when an approved issue (issue.md) needs its technical plan — produces the fused spec.md + tasks.md just-in-time, self-reviews them, then drives the issue pipeline: implement, quality gate, runtime verification, PR, review to lgtm, learnings. The issue owner's skill.

Review a branch diff (or any diff/files pointed at) with specialized find-only subagents — rubric+conventions (the universal coding standard plus the project's conventions), issue-conformance (the issue's AC-N and its runtime-verification results), and documentation-consistency (stale or missing docs after the change) — merged into one plain-text verdict that reaches lgtm only when every lane is clean. Classify findings (blocker/suggestion/nitpick/question). Portable: no dependency on a native review tool. Trigger on 'review this branch', 'code review', 'review the diff', 'review again', or the delivery step of the issue pipeline.