Skills for direct user invocation via slash commands. These are typically
invoked explicitly by the user (e.g., /nm, /proceed) rather than
automatically by the agent.
| Skill | Purpose |
|---|---|
capture-context |
End-of-session audit and handoff: catalog completed work, surface unresolved items, propose durable capture destinations, and ensure nothing from the session is lost before exit. |
extend-document |
Safely append or integrate new content into an existing document without losing original content. Use when adding research findings, session notes, or new sections to an existing .md file. |
for-later |
Cheap, non-disruptive capture of a topic the user wants to defer until later in the session. Records without analysis, continues in-flight work uninterrupted, and resurfaces the item at natural review points for disposition. |
retrospective |
End-of-session workflow analysis: timeline of errors and retries, skill and tool effectiveness audit, project documentation gaps, external knowledge dependencies, permission anomalies, and concrete improvement recommendations. |
The name prompting/ originally described "skills invoked via slash
commands" — which is true of the skills here but also true of almost any
skill. The three skills currently in this collection share a more
specific theme: they're generic cross-cutting workflow tools that
don't belong to any specific domain (coding, consulting, etc.). If this
repo ever reorganizes its collections around subject matter rather than
invocation style, the contents of prompting/ would likely move into a
collection named something like workflow/ together. Until then, this
is the general-purpose home for non-domain-specific workflow skills.