dsh-skill-dossier

You have downloaded dozens or even hundreds of skills, and the directory only shows a name and a one-line description. People flip through source files to jog their memory; models just guess. Whether a skill actually ran successfully, how long its entry occupies system context, or how stale it is—none of that is visible in a plain listing.

dsh-skill-dossier builds a readable dossier for every skill: direction category (10 types), scope, capability boundaries, applicable scenarios, source annotation, call statistics, real-world success and failure derived from DSH tools/result events, estimated token cost of the persistent index entry, and a freshness review queue. Three model-side tools—skill_dossier (read by name), skill_archive (write), skill_review (list pending reviews)—let an agent decide whether to load the full text of a skill before committing to it. The report panel is strictly read-only: it renders the daily briefs in your workspace into daily, weekly, and monthly views. Weekly and monthly views include a contribution chart where each cell is one day, color depth maps to total progress entries, and any fallback is labeled with the actual source date rather than pretending today has a record.

Built for engineers running a DSH web profile with a large skill set who want both AI and humans to understand each skill boundaries at a glance, and for teams with a daily-brief habit looking for a zero-config weekly or monthly rollup without spinning up a scheduler or an exporter.