Source-backed skill library
Every package links to its source and labels XSkills adaptations. Creator badges require authorship or review of the exact package.
Browse free skills
Choose a category, compare the outcomes, and open the complete skill before copying anything.
20 skills
Turn a measurable optimization problem into a bounded agent loop that proposes one change, runs the same evaluation, keeps improvements, and records every result.
You have one primary metric that can judge whether a change improved the system.
Turn a repeatable coding task into a bounded agent loop that discovers work, acts in isolation, verifies the result, records state, and stops safely.
You repeat the same bounded coding task often enough to automate it.
Follow Finn’s original six steps from paid-before-build validation through ICP-specific content and the long game.
You need paid demand evidence before building more product.
Turn parallel Claude Code sessions into reviewable pull requests through plan approval, isolated ownership, reusable commands, shared instructions, and end-to-end verification.
You have several bounded Claude Code tasks that can run locally or in the cloud.
Convert recurring agent failures into concise repository instructions or fast verification tools, then prove the improved harness prevents the mistake.
You see an agent repeat wrong commands, APIs, conventions, or unverifiable behavior.
Decide which tasks deserve a parallel coding agent by separating disposable research, reusable explanations, low-risk maintenance, and implementation with a reviewed specification.
You have more possible agent tasks than you can review at once.
Eric Provencher’s concise Codex skill for delegating substantial work across focused agents while the coordinator remains available and integrates the result.
You have a substantial, large-scope task that should be coordinated across focused agents.
Turn a curated watchlist into a persistent daily editorial queue that verifies evidence, maps existing coverage, packages three differentiated angles, and waits for a human decision.
You repeatedly monitor trusted sources for ideas and need more than a noisy link feed.
Convert a research question into a dependency-aware agent graph with independent research branches, a separate skeptic, evidence-gated synthesis, and manual approval before delivery.
You have a research task with genuinely independent evidence streams.
Rebuild a complete landing page by clarifying copy, preparing targeted design skills, building special effects separately, supplying real business context, and analyzing performance and SEO.
You need to test a materially new landing-page direction without a long handoff cycle.
Drive one complete repository goal through decomposition, parallel implementation, integration, live end-to-end verification, review, and a ready commit without stopping at partial progress.
You have one complete software task or specification with acceptance criteria.
Apply Theo's source-era model rankings and routing defaults, use the named Codex execution paths, judge the result, and escalate when the first route misses the bar.
You have several available models or agent harnesses with different strengths and costs.
Turn one of KingBootoshi's recurring weak-test patterns into a custom ESLint rule that prevents an agent from finishing through that pattern.
You see the same low-value testing pattern recur in agent-authored changes.
Reconstruct a prior coding-agent session from an authorized local transcript and the current workspace, classify unfinished work, and continue from the verified stopping point.
You are continuing work from a Claude Code session in another coding agent.
Pair an agent's miss with the user's corrective action, distill the pair into readable procedural memory, review it separately, and make it available to the next matching run.
You operate an agent in a domain where automatic right-or-wrong scoring is incomplete.
A source-grounded Fable workflow for broad-goal building, independent grading against real output, and repeated biggest-gap improvement. XSkills adds authorization, privacy, budget, and consequential-action gates.
You have a creative or engineering artifact with a concrete bar tied to inspectable real output.
A narrowly scoped reconstruction of Dax Raad’s demonstrated browser-control, HAR-capture, and derived-client sequence for authorized read-only retrieval. It excludes ordering, payment, account mutation, and access-control evasion.
You are authorized to perform a specific read-only browser retrieval task and inspect its network activity.
A planning and review skill for Weaver SDK widget-runtime work: understand before building, use receipt-backed tripwires, and expose budget failures as actionable agent-facing diagnostics.
You are planning or reviewing Weaver SDK widget-runtime or stack changes.
A consent-scoped workflow for turning authorized clothing photos into a reviewable wardrobe collection, proposing traceable outfit combinations, and optionally preparing GPT-image outfit concepts.
You authorize a defined photo set for wardrobe extraction.
A source-faithful implementation or review framework for a claimed YOLO, OpenCV, and Python match-video CV pipeline. XSkills adaptations add authorization, calibration, uncertainty tracking, and conservative metric rules.
You have authorized match footage and need a video-CV implementation or review.