Claude Skills by Silence-view

Proactive AI assistant mode for Claude Code targeting Opus 4.7. v3.0 overhaul makes Jarvis ACTUALLY useful from day one: lowered anticipation threshold (3 observations instead of 10), session-start awareness cards showing your top patterns, plugin monitors auto-arm without explicit invocation, first-run onboarding wizard. Anticipates next steps, auto-invokes relevant skills, and learns user workflow patterns via ROSETTA + auto-memory dual-layer. Three phases: Observation (1-3), Anticipation (...

Use when user explicitly invokes "/llm-language" or says "llm-language", "optimize prompt", "re-engineer prompt", "massima precisione", "maximum quality", "ottimizza il prompt". This skill re-engineers the user's prompt through multi-agent debate with scientific optimization, ROSETTA.md + auto-memory dual-layer persistent memory, mandatory codebase awareness, web deep research, full skill access, AND v4.1 skill-level meta-refinement (audits relevant user-owned skills, applies safe fixes autom...

Deep audit and refinement of third-party skills in the user's Claude Code ecosystem. Scores each skill on 10 quality dimensions, proposes non-destructive fixes (deprecated params, emphasis markers, missing frontmatter), applies safe fixes automatically, and reports risky changes for user approval. Scoped to user-owned skills ONLY (never modifies marketplace plugins or Claude Code built-ins). Use when: "/llm-language:refine-skills", "audit my skills", "fix my skills", "improve skill quality", ...

Auto-research new Claude Code features, prompting innovations, memory modalities, and best practices (targeting Opus 4.7 era). Searches academic papers, Anthropic changelog, YouTube, and community resources to identify updates that should be integrated into llm-language. v2.0 adds parallel agent dispatch, YouTube research via yt-learner, agent framework comparison, memory modality tracking, and model-specific migration checks. Use when: "/llm-language:update", "check for updates", "update llm...