Expo Skills – Official AI Agent Skills for Expo

Evidence shows: README explicitly states automatic telemetry is off by default and only sends anonymous usage events (skill name, platform, hash of random install ID) when enabled, with no code, prompts, file paths, or personal data. This shows basic awareness of data flow transparency and sensitive data handling. However, no detailed permission list or least-privilege principle is stated, nor is it clear whether user confirmation is required for side effects like network requests or file modifications. Dependency security lacks a dependency list or security audit. External effects may include triggering builds or deployments, but user confirmation is not specified. Rollback mechanisms are absent. Source attribution is unclear because the publisher is unverified despite claims of being official. Deductions: lack of explicit permission declarations, user confirmation mechanisms, dependency security info, external effect explanations, rollback mechanisms, and verified source attribution.

Evidence shows: README and skill descriptions are structurally consistent, with clear grouping of skills, but no internal consistency checks are provided. Dependency availability relies on external services (EAS, Expo CLI) without guarantees. Failure messages are not addressed; no error handling or user prompts are described. Deductions: lack of dependency availability assurances and failure handling mechanisms.

Evidence shows: README clearly identifies target audience (AI coding agents) and multiple use cases (build, deploy, upgrade, debug), with installation guides and example questions. Capability boundaries are defined by distinguishing framework vs services skills and free vs paid. Trigger precision is supported by skill descriptions for agent selection, but detailed trigger conditions are not provided. Environment fit covers multiple agents (Claude Code, Codex, Cursor, etc.) but compatibility differences are not explained. Deductions: insufficient trigger conditions and environment compatibility details.

Evidence shows: README provides clear information architecture including skill lists, installation guides, update methods, FAQ, and telemetry notes. Installation notes are detailed, supporting multiple methods. Naming is stable, but no version history or changelog is provided. Known limitations only mention experimental skills may change or retire, but other limitations are not listed. License is MIT, clear. Versioning lacks version numbers or changelog. Maintenance responsibility is claimed as official but publisher is unverified, so responsibility is unclear. Deductions: lack of version history, changelog, and clear maintenance responsibility.

Evidence shows: Skills aim to provide focused Expo knowledge to help agents apply Expo guidance correctly, with outputs directly usable in projects. Marginal value is provided by official knowledge reducing agent errors. Cost-benefit is low installation and usage cost via CLI or plugin, but no performance or resource consumption data. Deductions: lack of empirical output quality and quantified cost-benefit.

Evidence shows: README claims skills are based on Expo docs and CLI but provides no specific citations or links. Cross-source corroboration is absent. Fact-inference separation is partially addressed by distinguishing skills from official docs, but it is not explicit which statements are facts vs inferences. Deductions: lack of traceable citations and cross-source verification.