Agency — Productivity skill for Claude Code
Run a B2B task across three bounded agents (Analyst / SDR / Auditor).
How to install Agency
Installs to ~/.claude/skills/liminal-space-io-liminal-agents-agency/SKILL.md
mkdir -p ~/.claude/skills/liminal-space-io-liminal-agents-agency && curl -fsSL https://raw.githubusercontent.com/liminal-space-io/liminal-agents/HEAD/commands/agency.md -o ~/.claude/skills/liminal-space-io-liminal-agents-agency/SKILL.md Restart Claude Code, or start a new session, for it to be picked up.
What Agency does
description: Run a B2B task across three bounded agents (Analyst / SDR / Auditor). The in-lane agent produces work; the others refuse and name the correct agent. Refusal is the feature. argument-hint: ""
Invoke the `agency` skill and follow its SKILL.md flow exactly: pass the task to the orchestrator, render all three agent outputs (lead with the in-lane agent, show refusals verbatim), and offer to store a correction.
Task: $ARGUMENTS
Alternatives in Productivity
- Om Correct — Sweep a corrected fact through the vault — finds every note restating it, applies the correction at the single 4.6k ★
- Schedule Remove — Remove a scheduled task by ID or name 510 ★
- Quick Start — Building From Source — TL;DR: A full rebuild requires Bun (not Node.js) for its compile-time intrinsics (feature(), MACRO, bun:bundle 175 ★
Full documentation available on GitHub
View Source RepositoryRelated Skills
Agent Reach
Route a task to the right specialists across the full agency-agents roster and execute it
Agents Reach
Route a task to the right specialists across the full agency-agents roster and execute it
Three Axes Log
Append a completed-task entry to the repository's BITACORA.md work log, compacting older entries when the file
Subtask.Prompt
Delegate a bounded child task to a specialist agent and get a structured result back — use for isolated, well-
Gaia Plan
Plan a complex feature using GAIA's task-orchestration pattern, structures the work into fresh-context subagen
Phased Dev Workflow
A Claude Code/Codex CLI skill for feature and bug-fix work: it breaks large changes into reviewable phases, as
Related Agents
Liminal Analyst
The Liminal Analyst — diligence, competitive teardowns, market research, data enrichment. Bounded; refuses out
MCP Safety Reviewer
Reviews MCP tool designs for blast radius, three-mode design (read_only/draft_only/approved_execute), refusal
Namer
Invents a name for a persona that has none, so the others at the table can address them. Reads nothing and wri