RAD-Claude-Skills

Review my code, code review, is this ready to ship, check for bugs, security audit, review this PR, pre-merge check, is this safe to deploy, check code quality. The specialist lanes and memory that the built-in /code-review doesn't have: it invokes the built-in engine for general bug-finding, then adds a mechanical hallucinated-imports validator (with a vendored slopsquat denylist; Python/JS/TS/Rust/Go), an AI-slop lane, framework IDOR heuristics, a Supabase/Firebase BaaS-RLS security lane, and a deterministic findings index that tracks CR-NNN IDs across runs (new / recurring / resolved). Fixes apply to the working tree only — never commits.

This skill should be used when the user says "adopt this repo", "bring this repo onto the doc model", "this repo predates the system", "onboard this existing project", "migrate this repo's docs", or when startup found an established repo without the doc model. Brownfield archaeology, read-only toward code: reads git history + docs + code, drafts AGENTS.md/handoff/decisions plus an architecture.md module map from evidence (verifying build/test commands by actually running them), triages every existing doc onto the shelf (keep / merge / satellite / archive — proposing each move, never silently deleting), asks only what evidence can't answer, stamps dormant repos, runs fit-out, and hands plan reconstruction to /rad-planner:rescue. Never changes code.

This skill should be used when the user says "repo-align", "align the repo", "clean up the repo", "check for drift", "is the repo in good shape", "find contradictions", "what's gotten messy", "doc cleanup", "are my docs consistent", "tidy up the docs", or when startup's trust report went red. The deep, opt-in alignment pass: mechanical scans (drift, freshness, structure drift, vocabulary, size budgets) plus judgment passes — a doc router that reads off-shelf files and proposes where their content belongs, a routes-vs-api.md and modules-vs-module-map diff, and an on-request rules audit of AGENTS.md. Proposes before acting on every judgment call; preserves history with git mv; never silently merges, deletes, or archives.

This skill should be used when the user says "set up this repo", "bootstrap this project", "new project setup", "scaffold the docs", "initialize the repo model", "get me started right", or when startup recommended it on a fresh repo. Builds the container for the doc model in a new or nearly empty repo: an AGENTS.md skeleton with the stamped doc-model block, the CLAUDE.md @AGENTS.md shim, docs/ + docs/archive/, and a handoff stub. Creates no shelf or initiative docs (their triggers haven't fired), runs no fit-out, never invents product content, and ends by recommending /rad-planner:plan. For an existing repo with history and docs, use /rad-repo-manager:adopt instead.

This skill should be used when the user says "ship", "ship it", "close out and push", "wrap and ship", "wrap it up and push", "send it", "close-out", "commit and push everything", or "end the session and push". The one-command close-out chain: align-lite → quick wrapup → reviewed staging → repository contract gate → conventional commit → push → deploy-verify (only when AGENTS.md declares a deploy: target) → stray branch/worktree sweep → one-line report. Invoking /ship IS the owner's authorization to commit and push — no re-asking. Target: under three minutes including deploy verification.

This skill should be used when the user says "startup", "start session", "orient me", "where did we leave off", "catch me up", "what's the state", "session briefing", or "what was I working on". Fast, read-only session orientation — read the L0/L1 docs + scoped instruction map + git state, run the cheap mechanical scans, and surface a trust report (one line per managed doc, measured in commits-behind) plus the next task from the handoff. Recommends /rad-repo-manager:repo-init on a fresh repo, /rad-repo-manager:adopt on an established un-managed repo, and /rad-repo-manager:repo-align when the trust report goes red. It does not scaffold, audit deeply, clean, write, or change anything.

This skill should be used when the user says "wrapup", "wrap up", "end of session", "save state", "handoff", "leave a clean stopping point", "I'm done for now", or "before I close". Quick by default (target under a minute): overwrites docs/handoff.md (≤60 lines) from git evidence — not chat memory — carrying the Deferred ledger forward, then asks exactly two questions: any decisions settled this session (append to docs/decisions.md)? anything blow up worth remembering (append to docs/lessons.md)? The deeper reconcile pass — checking whether the session left plan/prd/AGENTS.md stale, plus structural signals in the session diff (routes → api.md, modules/deps/migrations → architecture.md, visual decisions → design.md), drafting per-edit-confirmed fixes — runs only with --full (weekly use). No status/roadmap files, no auto-commit or push, never runs tests.

Delegate coding, debugging, analysis, or bulk mechanical work to the local OpenAI Codex CLI (GPT-5.5) running as a non-interactive sub-agent. Use when the user says "use codex", "ask codex", "have codex do/fix/build/implement", "get a second opinion from GPT/codex", when routing bulk mechanical work (migrations, spec implementations, data transforms) per standing routing rules, or whenever Claude Code needs to run the Codex CLI programmatically. For reviewing a diff/branch/commit, prefer the rad-codex:review skill.

Cross-model code review — have the local OpenAI Codex CLI (GPT-5.5) review the working diff, a branch delta, or a specific commit and return findings for Claude to verify. Use when the user says "codex review", "have codex review this", "get a second set of eyes / second opinion on this diff", "cross-model review", or before shipping high-stakes changes when an independent GPT perspective is wanted alongside Claude's own review.

Delegate work to the locally-installed, authenticated OpenAI Codex CLI (`codex exec`) as a subagent from Claude Code. Use when the user says "use codex", "ask codex", "get a second opinion", "have codex review this", "compare codex vs claude", or wants to offload a large / context-heavy coding task (multi-file refactor, whole-repo analysis, long research) so it doesn't bloat this conversation's context. This is the lightweight bash companion to the official codex-plugin-cc MCP plugin — prefer this for one-off shell delegation, prefer the plugin for tighter tool integration.

This skill should be used when the user says "plan my project", "create an implementation plan", "help me architect this", "I need a plan before coding", "let's plan before we build", "plan this feature", "break this down into tasks", "map out the work", "create a project plan", "plan this app", or wants a structured, risk-first implementation plan before writing any code. Also trigger proactively when a user describes a non-trivial project or feature and appears ready to start coding without a plan. Runs a structured discovery interview, builds a release map (MVP/Beta now, V1 next, end goal later), produces a single plan file (`docs/plan.md`), and ends by routing the structured leftovers of planning (settled decisions, cut ideas, architecture seed) onto the repo's doc shelf. Can draft `docs/prd.md` from the interview answers with per-section confirmation when none exists; changes to an existing PRD surface as an update-prompt for the user to apply.

This skill should be used when the user says "replan", "update the plan", "re-baseline the plan", "the plan is out of date", "we shipped the MVP — what's next", "mark this milestone done", "the plan doesn't match reality anymore", "pull the next version into the plan", "adjust the plan", or when an existing `docs/plan.md` has diverged from what actually got built. Evidence-based plan update: reads git history + docs/handoff.md to mark shipped work (history preserved, never deleted), re-baselines remaining tasks, gives every parked idea in docs/ideas.md a fair hearing, and pulls the next release-map horizon into detailed milestones when the current one is done. Re-lints and runs a single risk pass; nothing is final until the user approves.

This skill should be used when the user says "rescue this project", "this repo is a mess", "help me out of this", "I don't know where this project stands", "get this back on track", "untangle this project", "I abandoned this and want to restart", "figure out what state this is in", or has an existing project with unclear status, missing or garbage docs, and no trustworthy plan. Runs project archaeology (evidence-based state assessment from code + git, no code changes), reconstructs intent through the discovery interview, hands structure work to rad-repo-manager's adopt, drafts a PRD from the answers if none exists, and produces a fresh release-map plan from wherever the project actually is. It assesses and plans — it never fixes, runs, or deletes code.

Let's brainstorm, I need ideas, brainstorm with me, what should I build, generate ideas, give me alternatives, I don't know where to start, SCAMPER, six thinking hats, reverse brainstorm, how might we, creative unblock — facilitated ideation on any topic (software, business, content, travel, creative, personal), from blank slate to refining an existing concept. Runs as a quick pass (a few turns) or a full facilitated session. A specific technique can be requested by name as a mode (e.g. "scamper", "six-hats", "reverse", "hmw", "unblock") and runs inside the session frame.

Design this, create a spec, design sprint, architecture for this, write a design doc. Post-ideation: architecture, components, data flow, error handling, testing. Not for ideation — use brainstorm-session for that.

Evaluate these ideas, which idea is best, help me prioritize, compare these options, rank these, how do I choose. Structured evaluation via Impact/Effort Matrix, Assumption Mapping, Pre-Mortem, JTBD, weighted scoring.

Static analysis pass over JSX/HTML/CSS source for WCAG 2.2 AA failure patterns. Does NOT run axe-core, does NOT measure real contrast, does NOT test runtime behavior — pair with a11y-testing for that. Use when the user asks for an accessibility review, a11y check, WCAG review, "is this accessible?", "does my site meet WCAG?", "check my component for accessibility", "review accessibility of", "check for a11y issues", "fix accessibility issues", or any request to scan a web page, component, or codebase for WCAG 2.2 AA failure patterns. Findings are tagged by detection confidence; the report does not produce a Pass/Fail verdict because static analysis cannot defensibly produce one.

This skill should be used when the user says "write a prompt", "prompt engineering", "system prompt", "optimize my prompt", "create instructions for", "context engineering", "CLAUDE.md", "few-shot examples", "write me a Midjourney prompt", "create a GPT prompt", "prompt for Claude Code", "agentic prompt", "tool description", or needs to create, improve, or optimize prompts for any AI platform. Covers prompt writing, system prompts, tool descriptions, agentic workflows, prompt chaining, and cross-model migration. Use for any prompt-related task, even simple ones.

This skill should be used when the user says "hemingway bridge", "wrap up session", "save my progress", "where did I leave off", "session handoff", "end of session", "capture what I was working on", "pick up where I left off", "PARA handoff", "what should I do next session", or wants to capture session state in a PARA-aware format before stopping work. Integrates with rad-repo-manager's wrapup/startup cycle.

This skill should be used when the user says "organize my notes", "PARA method", "second brain", "set up PARA", "I'm overwhelmed with files", "productivity system", "knowledge management", "where does this go", "classify this note", "my system is broken", "digital organization", "clean slate", "project list", or wants to build, fix, or maintain a PARA-based knowledge management system. Covers PARA setup, classification, system diagnosis, and review workflows.

Full brainstorming suite: facilitated ideation with proven methodologies, creative unblocking, structured evaluation, idea stress-testing, and design sprint for turning ideas into specs.

Full SEO suite: site audit, competitor intelligence, keyword/content strategy, AI visibility optimization, and schema markup. Proactively analyzes sites and produces scored reports.

Use when the user asks to write, draft, compose, improve, polish, edit, rewrite, review, critique, proofread, or AI-audit prose — or to analyze and learn their writing voice. Covers 9 domains (email, blog, web copy, business reports, research/academic, presentations, prose/essay, technical docs, social media), AI-pattern avoidance organized by durability tier, and voice profiling with EMNLP-2025-grounded honesty about idiosyncratic-voice limits.

Convene a council, ask the council, debate this, get a panel to critique, run this by a council of experts, council of advisors, devil's advocate panel, red-team this decision, what would a panel say, multi-perspective critique, stress-test this decision, get diverse expert opinions and a recommendation. Convenes 3–5 cognitive-framework advisors that debate any decision — repo plans, website/UX designs, product or codebase critiques, marketing plans, strategy choices — then vote and return one ranked, confidence-rated recommendation with dissent preserved and a single concrete next step. Use when a decision genuinely merits several opposed expert lenses; not for quick single-answer questions or line-by-line code review (use rad-code-review for that).

AEO, AI visibility, AI search optimization, generative engine optimization, LLM seeding, AI citations, answer engine optimization, brand presence in ChatGPT/Perplexity/Google AI Overviews. This skill optimizes the user's OWN content structure and distribution for AI-extractability. It does NOT measure actual AI citation rates — that requires direct AI-platform API integration (Path B).

Broken links, 404 errors, dead links, redirect chains, link checker, fix links. Scans the user's codebase for broken internal/external links, redirect chains, mixed-content issues, and unlinked brand mention opportunities. Real code work — no Path B dependencies.

Competitor analysis, competitor SEO, who ranks for, competitive audit, compare my SEO, competitor gap. Covers content gaps, technical SEO comparison (observable signals), SERP feature ownership, AI citation pattern observation, and qualitative link opportunity mapping. Does NOT report numerical backlink counts, domain authority, organic traffic, or actual AI citation rates — those require a backlink / traffic / AI-platform MCP (Path B).

Content strategy, content gaps, editorial calendar, topical authority, content plan, what should I write about. Covers topical authority mapping, gap analysis, brief generation, and 12-week editorial calendar. Most of this skill is strategic reasoning over observable content + WebSearch — no Path B dependencies except for Phase 3 traffic scoring (requires GSC/GA data, which the user can paste in or skip).

Fix my SEO, fix this SEO issue, fix missing meta, fix broken links, fix page speed code-level risk, fix schema — user has a specific SEO problem to remediate. Routes to specialized skills and applies fixes. For diagnosis without fixing, use full-seo-audit instead. Fixes that require numerical measurement (e.g., "fix my LCP to 2.5s") need a Path B MCP to verify — surfaced in the route.

Audit my SEO, run an SEO audit, SEO score, SEO health check, full SEO analysis — comprehensive multi-category request. Performs a scoped 6-phase audit over the user's codebase plus a user-provided URL list (typically 5-10 URLs). Covers technical SEO, on-page SEO, content/E-E-A-T, schema, internal linking, and AI-extractability. Does NOT crawl the entire live site, does NOT measure Core Web Vitals numerically, does NOT infer backlinks — those require Path B MCP integrations. For single-area requests, route to the specialized skill instead.

Keyword research, find keywords, what keywords should I target, keyword analysis. Covers seed discovery, expansion, intent classification, qualitative difficulty assessment, and topic clustering. Produces 150-300 prioritized keywords. Does NOT return numerical search volumes or difficulty scores — those require a keyword-data MCP (DataForSEO / Ahrefs / Semrush — Path B). Instead, uses observable SERP signals for relative assessment.

Link building, backlink strategy, link acquisition, outreach plan. Covers linkable asset identification, tactic selection, and 12-week outreach playbook with templates. This is a tactics generator — it does NOT measure the user's backlink profile, domain authority, referring domain counts, or competitor backlink gaps numerically. Those require an Ahrefs / Majestic / Semrush / Moz MCP (Path B). The user can still hand-paste exported data from those tools if they have it.

Optimize this page, title tags, meta descriptions, heading structure, E-E-A-T, on-page SEO, content quality score — single page optimization. Covers title tags, meta, headings, keyword usage, E-E-A-T signals, and image alt text. Real code transforms over the user's content — no Path B dependencies.

Add schema, structured data, JSON-LD, rich snippets, schema markup, generate schema. Generates valid JSON-LD, validates existing markup, and identifies missing schema opportunities. Real deliverable — Claude reliably produces valid, typed JSON-LD. No Path B dependencies.

SEO report, generate audit report, summarize SEO findings, SEO dashboard, SEO roadmap. Aggregates findings from completed audit skills into a prioritized report. Runs AFTER audit skills. Produces scores, executive summary, and 30/60/90-day roadmap. Honestly surfaces which categories were measured vs. require Path B MCP integrations.

Code-level technical SEO audit: crawlability, robots.txt (including the AI-bot allow/block matrix and llms.txt), sitemap issues, indexation, redirects, canonicals, JS rendering, security headers, URL structure, and page-speed code-level risk factors. Does NOT measure numerical Core Web Vitals (LCP/CLS/INP) — those require a Lighthouse/PSI MCP (Path B). Does NOT crawl whole sites — operates on the user's codebase plus a provided URL list.

This skill should be used when the user wants to author agentic loops or goal/completion conditions — "write a loop prompt", "loop engineering", "set up a ralph loop", "run Claude overnight", "write a /goal condition", "goal mode", "goal engineering", "stop condition for my agent", "completion criteria", "definition of done for the agent", "long-horizon task setup", "keep the agent working until", "make Codex run until tests pass", "PLANS.md", "harness for a long run", "my agent stops too early", "my agent never finishes", or wants Claude Code /loop, /goal, Stop hooks, or Codex Goal Mode set up well. Covers loop prompt authoring, goal condition design, the four-file long-horizon scaffold, and anti-reward-hacking hardening. For ordinary single-shot prompts, use prompt-engineering instead.

This skill should be used when the user says "decompile this prompt", "reverse engineer this prompt", "analyze this prompt", "break down this prompt", "what does this prompt do", "adapt this prompt for", "simplify this prompt", "split this prompt", "migrate this prompt", "why does this prompt work", "explain this system prompt", "what techniques is this using", "convert this prompt from X to Y", "what framework is this prompt using", "optimize this prompt", "look at this prompt", or pastes an existing prompt and wants it analyzed, adapted, simplified, or understood. Covers prompt anatomy, technique identification, structural analysis, cross-platform migration, and weakness mapping.

This skill should be used when performing operational actions against a Coolify instance using MCP tools — deploying applications, checking deployment status, viewing logs, diagnosing errors, managing environment variables, restarting services, checking server health, or performing any hands-on Coolify management task. Trigger when: "deploy my app", "check Coolify status", "show me Coolify logs", "what's running on Coolify", "restart my app on Coolify", "check server health", "why is my deploy failing", "list my Coolify apps", "add env var to Coolify", "Coolify deployment status", "roll back my deploy", "check my databases", "what resources are on my server".

This skill should be used for a quick health/status overview of a Coolify instance — a one-shot "is everything up" dashboard. Trigger when: "coolify status", "is everything up", "how's my Coolify doing", "status check", "give me a Coolify overview", "anything down?", "morning check", "is my server healthy", "what's the state of my deployments". For deep diagnosis of a specific problem, use coolify-troubleshoot; for performing actions (deploy, restart, env vars), use coolify-actions.