qui-context-optimizer

Advanced context management with auto-compaction and dynamic context optimization for use with SkillBoss API Hub LLM services via /v1/pilot. Features intelli...

This skill appears to implement a context pruner for SkillBoss and legitimately needs SKILLBOSS_API_KEY and embedding/token libraries, but there are a few red flags to weigh before installing: - Documentation mismatch: SUMMARY.md claims no external deps, but package.json and code require tiktoken and @xenova/transformers. Expect to run 'npm install' and for runtime model downloads. - Local storage: the skill creates an archive directory and index.json and writes pruned conversation content to disk (unencrypted). If your conversations contain sensitive data, choose a secure archivePath or disable archive indexing. - Logging/leakage: archive.store and the chat-logger log message snippets to stdout and are designed to send formatted logs to chat; these logs can reveal user content. Review/override the logger (provide a safe onLog handler) before enabling logToChat. - Dependency/runtime downloads: embedding model downloads (~tens of MB) and native builds (tiktoken) may occur; run install in a controlled environment and review network activity. Recommendations before installing: 1) Inspect package.json and the full lib/index.js to understand exactly what is written to disk and where (archivePath). 2) Run npm install in a sandbox or CI runner first to observe build steps. 3) Set SKILLBOSS_API_KEY to a scoped/test key, not a high-privilege credential, until you confirm behavior. 4) If you need confidentiality, disable archiveIndexing or change archivePath to a secure location and override logging to avoid sending content to chat. 5) If the documentation mismatch bothers you, reach out to the skill author or prefer a version whose README and metadata are consistent.