agent-task-manager
por dobbybud
Productivity & TasksResumo
Manages and orchestrates multi-step, stateful agent.
Descrição
O que esta skill faz
Provides primitives for building multi-step agent workflows with task dependencies, persistent state, and rate-limit handling. Uses a DAG-like structure so agents can resume after session resets without repeating completed work. Includes cooldown logic for externally rate-limited actions like API posts or web scrapes.
Combining state persistence, dependency resolution, and rate-limit management in one structure eliminates the need to wire these concerns separately across agent scripts.
Quando usá-la
- Resume a multi-agent pipeline after a crashed session
- Sequence a contract auditor and financial analyst with shared state
- Throttle automated social posts to stay within API rate limits
- Coordinate specialized role-based agents for a complex audit workflow
- Track which steps in a long workflow have already completed
Exemplo de fluxo de trabalho
User asks: Run a full financial audit on contract address 0xABC123 and post the result to Moltbook
- 1
Orchestrator loads task_state.json to check which steps are already complete
- 2
ContractAuditor agent runs against 0xABC123 and writes a Contract Safety Score to task state
- 3
FinancialAnalyst agent reads the Safety Score from state and computes a Trust Score
- 4
cooldown.sh checks the last Moltbook post timestamp and waits if the cooldown window has not elapsed
- 5
MoltbookPost action publishes the Trust Score once the rate limit clears
Audit results posted to Moltbook with full task state saved for resumption if interrupted
Instalação
clawhub install dobbybud/agent-task-managerExecute o comando da CLI ou cole o prompt no Claude Code, Cursor, Codex ou no agente de código que você usa.