t-benze
@t-benze
Public Skills
thread
by t-benze
Use this skill when the orchestrator invokes you for thread participation. Decide whether to reply, decline, or dispatch a task — all based on the thread context provided in your prompt.
start-task
by t-benze
Use this skill at the start of every task. Parses task_id, session_id, brief, and role_guidance from the prompt, executes the work, reports completion back to the daemon, and cleans up worktrees.
workspace-cleanup
by t-benze
Shared daily/manual own-workspace cleanup contract — bounded, Git-aware, non-force reclamation of registered non-primary task worktrees and their dependency caches, with an exact same-user process observation and authoritative terminal/history/peer joins.
jobs
by t-benze
Run an authorized script in the background; inspect its actual result.
manage-repo
by t-benze
Add, remove, or update a repository in your AgentDef configuration (org/agents/<name>.md frontmatter). Write a JSON file and call happyranch manage-repo --from-file to keep the invocation single-line.
dream
by t-benze
Use this skill only when the daemon starts a private scheduled dream invocation. It produces private learnings, KB candidates, and optional founder-thread output.
manage-agent
by t-benze
Enroll, update, or terminate an agent. Write a JSON file and call happyranch manage-agent --from-file to keep the invocation single-line. Enrollment requires founder approval.
todos
by t-benze
Use when you need to create a scheduled Todo for yourself from explicit founder or operator instruction. Never infer or proactively schedule future work. Scheduling is self-only and available to every valid in-org agent.
make-worktree
by t-benze
Use this skill before any git commit, git checkout, or file edit inside repos/<name>/. Read-only exploration does not need a worktree. Manages a per-task git worktree at .claude/worktrees/<task_id>/ on branch task/<task_id>. Activates a worktree-root guard that prevents accidental primary-checkout edits.
happyranch
by t-benze
Manage the HappyRanch AI tourism organization via the happyranch CLI — submit tasks, stream events, inspect audit logs, approve agent enrollments, operate the shared knowledge base, manage threads (compose/send/invite/archive/abort-replies), and record founder-resolved precedents. Use when the user asks about HappyRanch tasks, agents, threads, the daemon/runtime, the knowledge base, or the one-person-company system.
create-skill
by t-benze
Verified-agent custom-skill authoring from an active task session. Creates a standard_operational custom skill with server-verified task/session/agent/org provenance. The B2 record is created immediately and hidden by default until founder-configured eligibility makes it visible.
reflection
by t-benze
Use this skill when asked mid-thread to reflect on your recent work. Produces an opening report (recent tasks, learnings, frictions) posted as a thread reply, and captures learnings + optional KB entries via direct happyranch learning add / happyranch kb add calls at the end.
reflection
by t-benze
Use this skill when asked mid-thread to reflect on your recent work. Produces an opening report (recent tasks, learnings, frictions) posted as a thread reply, and captures learnings + optional KB entries via direct happyranch learning add / happyranch kb add calls at the end.
manage-agent
by t-benze
Enroll, update, or terminate an agent. Write a JSON file and call happyranch manage-agent --from-file to keep the invocation single-line. Enrollment requires founder approval.
manage-repo
by t-benze
Add, remove, or update a repository in your AgentDef configuration (org/agents/<name>.md frontmatter). Write a JSON file and call happyranch manage-repo --from-file to keep the invocation single-line.