Skills

Build, debug, or review Cloudflare Agents SDK applications using the agents package.

Read-only repository explorer for cold-start orientation, broad cross-file localization, or when a direct search failed. Skip it when the exact file or symbol is already named. Returns path:line citations only; its reads stay out of main context.

Restructure code while preserving behavior. Use for extraction, consolidation, ownership moves, or cleanup where verification must bracket structural edits.

Act on a Caveman learn report - review the ranked token sinks, apply cost-lowering fixes with per-edit consent, and report what those fixes returned. Use when asked to lower an agent's token cost, what caveman has saved, to trim a heavy CLAUDE.md, or to offload re-pasted context into cavemem.

Turn a Caveman optimization observation into an operator-chosen candidate with a paired baseline evaluation. Use when asked to inspect or evaluate a Caveman optimization report. Needs explicit approval.

Review docs/prose for Writing Guidelines compliance. Use when asked to "review my docs", "check writing style", "audit prose", "review docs voice and tone", or "check this page against the writing handbook".

Diagnose ambiguous failures before editing. Use for unknown causes, intermittent behavior, performance regressions, or investigations needing evidence-ranked hypotheses.

Build AI-powered software with TypeSafe: small units of AI intelligence you can use like programming primitives. Its System One models, including Jev, turn natural language and application state into typed judgments and probabilities that code can combine. Use when a feature needs programmable common sense, when brainstorming what AI could make possible in an app, or when an LLM prompt-and-parse step could become a structured decision. Applications include routing, ranking, extraction, verification, and int

Prove existing work meets acceptance conditions without expanding scope. Use for validation-only tasks, completion checks, focused gate runs, and last-mile proof.

Fix bugs and small behavior changes at the narrowest responsible layer. Use when regression proof, preserved surrounding behavior, and task-relevant tests matter.

Read-only review of Caveman Cloud evidence: cost, Cave Score, workflows, traces, latency, errors, routing, savings. Use when asked what Caveman found or where LLM spend goes.

Build feature work with high overbuilding risk. Use for new behavior, product slices, or integrations where repository reuse, strict scope, and an explicit stop condition matter.

Find and label every LLM workflow in the repository so Caveman Cloud groups spend by workflow instead of one bucket. Use for "discover workflows" or breaking LLM spend down by workflow.

Inspect Caveman Cloud's experiment lifecycle and block unsafe execution. Use when asked to start, approve, cancel, promote or roll back a Caveman experiment.

Implement or troubleshoot Cloudflare Email Sending and Email Routing integrations and their delivery configuration.

Implement reversible compatibility-safe transitions. Use for schema, data, API, protocol, configuration, or dependency migrations requiring rollback and preservation proof.

Wire a repository through the Caveman Cloud gateway so every LLM request is measured, with no behavior change. Use for "set up caveman" or adding LLM spend observability.

Use when user asks to "build a Solana dapp", "write an Anchor program", "create a token", "debug Solana errors", "set up wallet connection", "test my Solana program", "fuzz my Solana program", "deploy to devnet", "send a v1 transaction", "support larger transactions", "fix maxSupportedTransactionVersion", or "explain Solana concepts" (rent, accounts, PDAs, CPIs). Also for program architecture — state layout, reducing compute units, throughput bottlenecks, instruction naming — and quick on-chain lookups via

When the user wants to research, profile, or analyze competitors from their URLs. Also use when the user mentions 'competitor profile,' 'competitor research,' 'competitor analysis,' 'profile this competitor,' 'analyze competitor,' 'competitive intelligence,' 'competitor deep dive,' 'who are my competitors,' 'competitor landscape,' 'competitor dossier,' 'competitive audit,' or 'research these competitors.' Input is a list of competitor URLs. Output is structured competitor profile markdown files. For creatin

How to write modern Swift well — modeling with value types, Swift 6 data-race safety and approachable concurrency (@concurrent, main-actor-by-default, actors, task groups), protocols and generics (some vs any), API design, performance and ARC, Swift Testing, macros, and the modern language features agents don't know about yet. Use when writing, reviewing, or migrating Swift, or when a concurrency error, a hang, a data race, a retain cycle, or a performance problem needs fixing.

Audit a browser fingerprint for internal contradictions with the liarjs CLI - canvas, WebGL, WebGL2, WebGPU, audio, 220 fonts, WebRTC and timezone probes, scored against the TLS/HTTP/ASN view of the same request. Use when asked to run a browser fingerprint test, see what a fingerprint looks like, check canvas or WebGL fingerprint stability, compare a spoofed profile against a real browser, or find out whether a browser profile is self-consistent.

Gate a build on browser fingerprint regressions with liarjs - save a baseline scan as JSON, diff later runs against it, and fail the job when the consistency score falls below a floor. Use when asked to add a fingerprint or headless-detection check to GitHub Actions, GitLab CI or another pipeline, to catch a regression in a Chromium build or scraping harness before it ships, or to track how a fingerprint score changes across commits.

Check whether a Playwright, Puppeteer, Selenium or CDP-driven browser presents a coherent fingerprint, using liarjs as a library against a Page you already have - navigator.webdriver, HeadlessChrome tokens, worker versus main-thread identity, patched-API integrity, WebGL versus WebGPU GPU identity. Use when asked whether an automated browser looks like a normal one, when a headless setup or a stealth plugin's effect needs measuring rather than assuming, or when an assertion on fingerprint quality belongs in

Read a liarjs fingerprint report and attribute each failing check to the component that produced it - what the check id measures, whether the signal comes from the launch configuration, the page-modifying layer, the network path or the machine image, and which failures are inherent to headless or datacenter environments. Use when a fingerprint scan came back with a low score, or when a check id such as webdriver, worker-consistency, gpu-triad, native-integrity or tz needs explaining.