AgentOS 2026.7.30

Added

• Capminal Skills in the Skills hub — browse, inspect, and install allowlisted Capminal/agent-skills with publisher branding that falls back cleanly when the logo cannot be fetched (#144).

Changed

• Runtime dependencies carry upper bounds — a major release of a dependency can no longer land in an install that was resolved against the previous one (#153).

Fixed

• The skills block is now something the model can act on — the prompt budget no longer falls off a cliff that drops every description at once, names-only mode points at skill_list instead of one skill_view per skill, the block lives in the cacheable system prompt rather than a self-contradicting user message, and cron turns finally get the tools the block tells them to call.
• Asking for a skill that is not installed is no longer a dead end — skill_view says the lookup worked, names similar installed skills, and offers skill_search_community, instead of leaving models to invent failures like "returned error: 14" (#162).
• Reading a large skill no longer costs the whole skill — over [skills].max_skill_view_chars (new, default 10 000) skill_view returns the opening sections plus an index, read on with skill_view(name, section="<title>"): 43% fewer characters across a real install, 80–87% on the largest.
• A degraded skills render is now visible — each turn records skills_render_mode and skills_description_max_chars, and anything below full is logged as skills_filter.budget_degraded with what to change.
• Skill file links are POSIX on Windows — the index no longer offers references\api.md, which a model quotes back as a broken file_path.
• Skill cards stay in their grid tracks — fixed card height and contained text, so the hub grid no longer reflows as cards load (#135, #161).
• The Installed chip in the skill detail dialog reflects reality (#121).

Full changelog: v2026.7.29...v2026.7.30