md2wechat Skill: Markdown-to-WeChat Formatting & Publishing

Evidence shows: SKILL.md explicitly requires explicit user requests for side effects like upload, draft, publishing, and uses inspect readiness checks as a publish gate; config wechat-accounts is declared local-only and never prints WeChat secrets; Brand Profile creation/editing requires explicit user consent. However, reliance on external API services (MD2WECHAT_API_KEY) and image providers, whose availability, data flow, and privacy policies are not disclosed in this static review; the license is Source Available (based on BUSL-1.1), not a standard open-source license, with strong commercial restrictions, and the publisher is unverified. No code was executed, so covert credential handling or data transmission cannot be ruled out; these uncertainties lower the trust score. Deductions: external service dependencies and unverified execution behavior.

Evidence shows: SKILL.md provides detailed workflows, discovery commands, failure handling paths (e.g., doctor config show, layout validate), and expected outputs (e.g., PREVIEW_ACTION_REQUIRED). However, no test suites or CI coverage for the skill's key paths are visible; go.mod and CI workflows only show build and quality checks, not behavior tests. Thus, the happy path may be plausible, but edge cases, error handling, and failure feedback are thin. Static review cannot execute, so reliability cap is 10, hence the low score. Deductions: no test evidence, limited error-handling coverage.

Evidence shows: The skill targets WeChat article formatting and publishing for Chinese users, and the README provides Chinese documentation and examples, adapting to mainland-China context (though core may depend on overseas APIs). SKILL.md explicitly defines trigger conditions (e.g., detect markdown to WeChat formatting) and boundaries (e.g., no unrequested publishing). However, the skill relies on external APIs and image providers that may be unreachable from mainland China, and no offline alternatives are provided. Boundaries and triggers are clear, but environment-fit evidence is limited, hence the moderate-low score. Deductions: core functionality depends on overseas services, insufficient environment-fit evidence.

Evidence shows: SKILL.md is well-structured with sections (intent routing, discovery, config boundaries, etc.), the README provides installation guidance and docs links, and license and version info are clear (package.json version 3.2.0, CHANGELOG referenced). However, the license is Source Available, restricting commercial use, which may affect users; the publisher is unverified, and maintenance responsibility and update path are only claimed in README. Known limitations are mentioned in docs, but SKILL.md itself lacks version history or changelog. Deductions: non-standard license, incomplete maintenance responsibility info.

Evidence shows: The skill claims to complete Markdown-to-WeChat HTML conversion, draft creation, image generation, etc., with detailed command flows. But no verified output examples or comparative advantage data are provided; the core value (e.g., 40+ styles) is unverified in static review. Whether output is directly usable is untested, and marginal value over manual or alternatives is unclear. Static review cannot execute, so effectiveness cap is 7, hence the low score. Deductions: no verified outputs, insufficient comparative-benefit evidence.

Evidence shows: The repo has CI workflows and go.mod, but quality gates do not include tests for the skill's key paths; the README demo GIF is visual evidence but not executed. No third-party test reports or independent validation. Static review cannot independently reproduce, so verifiability cap is 5, hence the low score. Deductions: no audited evidence, insufficient test coverage.