claude-skills

Frontier-model-plans, cheap-model-executes delegation loop. The expensive model writes a decision-complete spec, dispatches implementation to a cheap lane (cheaper Claude subagent or Codex CLI), enforces an implementer/reviewer split with fresh-context adversarial review, and accepts work only by verified evidence — spending frontier tokens on judgment, not typing. Use when the user says "architect this", "spec it and delegate", "plan expensive execute cheap", "have the big model plan and a cheap model build", or invokes /architect <task>.

Get an independent, separate-model (OpenAI Codex CLI, flagship GPT tier) review of code — a just-shipped feature, a risky change, or a substrate you're about to build on. Use to complement your own Claude subagent audits with a genuinely different model that has different blind spots. A separate model catches what same-model reviewers miss.

The operating method Fable (Claude Fable 5) left behind — an evidence-first working loop plus the judgment layer for calibration, scope, and reporting. Use at the start of any substantial or ambiguous task, when unsure how to proceed, when work feels like flailing, or when the user says "work like Fable", "fable method", "how would Fable do this", or invokes /fable-method.

Validate every plugin in this marketplace for structural + registration consistency — valid plugin.json, name matches its directory, registered in marketplace.json and the README, SKILL.md frontmatter present, scripts executable and compilable, references over 100 lines carrying a Contents TOC, no pre-plugin-era script paths, and SKILL.md spines under 500 lines. Use when the user says "lint the marketplace", "validate the plugins", "check the skills repo", "are all my plugins registered", or invokes /marketplace-lint. Run it before pushing a new or edited plugin.

Cross-project forgotten-work radar. Sweeps every local git repo for uncommitted, unpushed, stashed, and never-pushed work, cross-references the context packs' open threads, and ranks what is most at risk of being forgotten — with one concrete first move each. Use when the user asks "what am I forgetting", "open threads", "loose ends", "what's at risk", "what did I leave hanging", or invokes /open-threads. (Distinct from /rehydrate, which restores ONE project's context, and /session-recap, which is a timeline of past sessions — this is the radar across everything.)

Address review comments on your own PR via an interactive walkthrough. Fetches unresolved comment threads, walks through them one at a time, lets you choose Fix / Push back / Defer / Skip per comment, dispatches an implementer subagent for each Fix, batches the fixes into one commit, posts replies (with commit SHA for fixes, reasoning for push-backs), and resolves the threads. Use when the user says "respond to PR comments", "address the review", "respond to feedback on PR 42", or invokes /pr-respond.

AI-powered PR review with contract verification, frontend and backend rules, inbox mode, and incremental re-review. Auto-detects project type and applies relevant rules. Use when user says "review PR 42", "review my PRs", "check my PR inbox", "pr review", or invokes /pr-review. This is the standard single-model review lane that posts to GitHub — for a multi-model panel use /review-board; unsure how much review a change deserves, /risk-gate first.

Quality-gated push pipeline. Auto-detects package manager (pnpm/yarn/npm) and runs available scripts (check-types, lint, test, build) in parallel before pushing. Auto-fixes failures in a retry loop, creates a PR, and polls CI. Use when user says "push", "push to remote", "ship it", or invokes /push.

Restores working context for a project after time away by reading its dated context pack, verifying claims against live git/filesystem state, and briefing with a visual status map plus the recommended next thread. Use when the user says "what was I working on", "resume <project>", "getting back to <project>", "catch me up", "back from leave/vacation/break", or invokes /rehydrate. This restores context to resume ONE project — for a visual timeline of past sessions use /session-recap; for the cross-project forgotten-work radar use /open-threads.

Multi-lens, multi-model review conductor. Audits a diff/PR/branch by spawning several scoped Claude review agents in parallel (security, concurrency, performance, API-contract) WHILE a separate model (Codex) reviews independently, then reconciles every finding into one cross-model table: convergence (high-confidence), net-new-from-Codex (verify), and sharpened disagreements. Use when user says "review board", "full review", "multi-lens review", "audit this with all the models", or invokes /review-board. This is the heavyweight lane of the review ladder — for a routine review that posts to GitHub use /pr-review; to spend this only where it pays, /risk-gate first.

Triages a diff/PR/branch by change risk and escalates only the risky parts to deeper review. Runs a cheap deterministic scoring pass (danger paths like auth/payments/migrations, change size, blast radius, churn) to bucket changed files Low/Medium/High, then escalates only HIGH-risk files to a heavyweight separate-model review (review-board or codex-second-opinion). Use when user says "risk gate", "what's risky here", "triage this diff", "is this safe to merge", or invokes /risk-gate. This is the cheap entry point of the review ladder — start here when unsure how much review a change deserves.

Generates a self-contained, browser-viewable HTML timeline of your recent Claude Code sessions — grouped by day, with key decisions, PRs, commits, files touched, active time, and a resume command per session. Parses ~/.claude/projects transcripts; defaults to the last 7 days (pass a number or "daily"/"weekly" to change). Use when the user wants a daily or weekly recap, asks "what did I work on", "summarize my Claude sessions", "session timeline", or invokes /session-recap. This is a visual record of what already happened — to restore working context and resume a project, use /rehydrate instead.

Closes a working session cleanly — repo left consistent, session logged to the knowledge base, every open thread named, context packs healed. Use at the end of any substantive session, before stepping away from unfinished work, or when the user says "wrap up", "hand off", "closing time", "done for today", or invokes /handoff.

Runs the weekly operator rhythm for an autonomous content/growth engine — automation health, queue integrity, what's rendered but unposted, product-truth enforcement, funnel status, and launch execution. Use when the user says "marketing check-in", "run marketing ops", "what should I post", "how's the engine", "launch mode", or invokes /marketing-ops.

Audits dependency-manifest / lockfile changes in a diff/PR/branch. Runs a zero-network offline pass over the diff (semver delta classification, pin/range loosening, lockfile-integrity changes, newly-added install scripts, typosquatting, transitive blast radius), then fans out per-ecosystem scanner agents that run the real tools (osv-scanner, npm audit, pip-audit, cargo audit, govulncheck) and maintainer-health checks, and reconciles a verdict per changed package. Use when user says "audit dependencies", "check this lockfile", "review these package bumps", "supply chain check", or invokes /dep-audit.

Template skill demonstrating the orchestrator -> parallel agents -> aggregate pattern. Splits a target (file glob or directory) into chunks, runs one worker subagent per chunk in parallel, and merges their findings into a single report. Use when user says "skill starter", "scaffold a skill", "run the starter audit", or invokes /skill-starter.

Authors a new Claude Code skill/plugin in this marketplace, the right way. Interviews for intent and trigger phrases, decides the shape (inline skill vs orchestrator + parallel specialist agents), scaffolds the files, writes them against a distilled authoring guide, self-reviews against an anti-pattern checklist, and registers the plugin in the marketplace + README. Use when the user says "create a skill", "scaffold a plugin", "new skill", "add a skill to the marketplace", or invokes /create-skill.

Adversarially hardens a design doc / spec / RFC BEFORE coding. Runs a pre-mortem (assume it already shipped and failed, enumerate why) plus parallel design-review lenses (problem framing, edge cases, failure handling, data integrity, rollout/rollback, observability, scaling, dependencies/blast-radius, security/STRIDE), then reconciles into a prioritized hardening report with an APPROVED/REVISE verdict. Use when user says "harden this plan", "review my design/spec/RFC", "pre-mortem this", "poke holes in this design", or invokes /plan-harden.

Finds the most important MISSING tests in a diff/PR/branch. Fans out gap-finder agents over the changed files; each traces codepaths (changed conditionals, error paths, boundaries) and ranks untested ones by risk = impact x likelihood, not raw coverage %. Optional coverage (diff-cover) and mutation-testing passes confirm "covered-but-unasserted" gaps. Produces a prioritized list of tests to add with proposed cases. Use when user says "test gaps", "what's untested", "what should I test", "coverage gaps", or invokes /test-gap.