ship’s manual · guide

Skills and growth

A durable agent needs more than ten generic prompt snippets. 1Helm ships complete operational playbooks and a supply chain for adding more without turning the user into a plugin administrator.

Focused assignments

Every resident starts with a seven-skill operational core plus playbooks selected by its channel template. The complete shared catalog covers outcome ownership, Skipper handoff, obligations, skill discovery, memory, research, email, calendar, contacts, messaging, documents, spreadsheets, PDFs, meetings, projects, personal operations, travel, finance, support, software delivery, GitHub, data, media, infrastructure, security, and self-hosting.

Each defines activation cues, execution, boundaries, retained state, and verification. A resident sees a compact inventory of its assigned skills, loads a full procedure on demand, and can ask Skipper for another catalog skill when a task expands.

Crystallize proven procedures

After a repeatable workflow succeeds, a resident can supply its complete activation cues, ordered action, authority boundaries, recovery, retained state, and verification procedure plus concrete completion evidence. Generic one-paragraph skills and unverified guesses are rejected. Skipper can also learn from a local file/folder, URL, or notes in a visible thread. Corrections and durable behavior guidance remain separate from raw chat.

External catalog

1Helm searches SkillsMD's public API directly and displays every result it returns. SkillsMD currently describes its records as GitHub repositories; 1Helm does not inflate that count into individual procedures or call the registry curated.

Inspection at install time

Choosing a result starts source inspection rather than browse-time censorship. 1Helm resolves an immutable GitHub revision, finds a bounded skill document, scans it, records its hash and provenance, and wraps clean content beneath runtime authority. Content that fails inspection is not silently activated.