_shared (Shared SDD References)

Evidence shows this skill is an internal shared reference, explicitly marked as not invokable, prohibiting invocation as a skill, which aligns with least privilege. Documentation emphasizes user confirmation (e.g., marking requires user confirmation) and limited permission operations; no overreach or data leakage risks found. However, missing explicit dependency security and external effect isolation disclosure lead to deductions.

Static review cannot execute key paths; documentation depends on external tools (e.g., Engram, gentle-ai) whose availability is not verified here. The docs provide detailed failure handling, but lack automated test coverage, limiting reliability.

As a shared library, its purpose and non-invokable boundary are clear, audience is specific (SDD skill developers), and trigger conditions are clear. The skill is not directly invokable and primarily serves as reference; no direct impact on mainland-China network, but core functions rely on Engram which may be inaccessible, requiring cautious assessment.

Documentation is well-structured with naming rules, examples, and reference docs, but lacks versioning details (only metadata version field), changelog, and explicit maintenance responsibility. License is MIT, but update path is not mentioned, hence medium score.

As internal reference, it mainly provides unified conventions to other SDD skills, reducing duplication and ensuring consistency. Static review cannot verify actual execution effect, and the marginal value of such reference docs is hard to quantify, hence lower score.

All documents are author-provided references with no executable tests or CI evidence, nor third-party verification, so only static review is possible with low confidence, resulting in very low score.