Agent workflows playbooks
Give investigations a clear brief, useful evidence, and an explicit outcome. Choose a scenario, then the decision you need to make.
Unclear Agent Brief
An agent receives an underspecified task, so the intended outcome, available evidence, boundaries, and human decision point are difficult to determine.
Missing Evidence
An agent recommendation or investigation lacks the source observations, artifacts, or comparisons needed for a human to evaluate its reasoning.
Contradictory Recommendations
Agents or workflow stages recommend opposing actions for a similar question, while their inputs, evidence, criteria, or decision boundary are not aligned.
Repeated Investigation
Agents repeatedly revisit an issue without adding evidence or narrowing the question, causing duplicate effort and unclear progress toward a decision.
Stale Context
An agent works from outdated files, records, prior findings, or task assumptions, so a plausible result no longer describes the current system or decision.
Oversized Scope
An agent task covers too many systems, questions, or outcomes at once, making evidence collection, progress, and review difficult to control.
Inconclusive Result
An agent cannot support a clear conclusion because evidence is incomplete, conflicting, unstable, or unable to distinguish the relevant hypotheses.
Missing Verification Plan
An agent task proposes or records a change without defining the checks, evidence, owners, or stopping criteria needed to judge its outcome.
Weak Handoff Summary
An agent handoff summarizes work without enough context, evidence, uncertainty, or next action for the receiving engineer or reviewer to continue confidently.
Unsupported Confidence
An agent expresses strong certainty without evidence, comparison, or uncertainty language that justifies the strength of its recommendation.