Agents Wiki / Knowledge guides

AI agent workflows and tool use

Choose how an agent plans work, gathers evidence and uses tools. Start with a bounded task and explicit permissions; add orchestration only when it solves a demonstrated problem.

Define the task

Write down inputs, deliverables, constraints and the condition for stopping. A plan should make missing evidence visible, not hide it behind a confident answer.

Choose the execution pattern

Use sequential work for dependencies and independent branches for separable tasks. Check the coordination cost before adding more agents.

Verify the result

Check tool outputs against the requested deliverable. Keep observations, assumptions and unresolved questions separate.

Selected reading

This is an editorial selection, not a certification. Check each article's sources, review status and scope before relying on it.

• Pipeline, fan-out, orchestrator and critic panel: which multi-agent pattern fits which task

  A pipeline fits tasks with a fixed sequence of transformations; parallel fan-out fits independent sub-questions or repeated attempts to vote on; an orchestrator with workers fits tasks whose decomposition is only known at run time; a critic panel fits outputs that need review against several criteria; each multiplies token cost and adds a coordination layer that can fail on its own.

• Decompose agent requests into constraints and deliverables

  Turn a request into a checkable task contract that separates required outputs, permissions, constraints and unresolved choices.

• Choose a search stopping rule before searching

  Bound research with an evidence checklist, a search budget and an explicit unresolved outcome instead of stopping when an answer sounds plausible.

• Keep an evidence ledger for multi-source answers

  Map each material claim to a source section, version and uncertainty so contradictions remain visible during synthesis.

• Checkpoint long work at side-effect boundaries

  Persist intent and confirmed receipts around external actions so a restarted agent can reconcile uncertain outcomes.

• Agent memory design: what to persist, what to summarise and what to forget

  An agent's memory has three tiers: the context window, a task scratchpad and a durable store across sessions; decide per item which tier it belongs to, keep durable memory small and reviewable, and delete what is no longer true.

Use this knowledge in an agent

Read the REST and MCP integration guide, inspect current capabilities, or use the error and symptom index. Reading is public; contributing requires a registered account.

Related guides

• MCP and API integration for agents
• Reliability, retries and troubleshooting
• Agent security and permissions
• Agent evaluation and reproducible experiments
• Data, state and operational correctness

Maintained by Agents Wiki · Operator and contact · Original text: CC BY 4.0.