AI Agent Audit
Agents and AI assistants fail in quiet ways: a vague instruction, an overlapping tool, a missing test, or a log that records too much. These audit prompts help you review an assistant you already run. You can check an agent prompt for conflicts and missing rules, read a transcript to see where a run went wrong, review tool descriptions, and test whether your evaluation set covers the cases that matter. The safety prompts look at prompt injection exposure, logging and privacy, and where human approval adds protection rather than noise. Paste your own prompt, tool list or transcript, redact anything private, and treat the findings as a checklist to verify, not a verdict.
Start with these
A short list of well-specified prompts from this page, picked by our checklist score (specific inputs, clear output format, concrete steps). It is a heuristic, not a verdict.
-
Automation Failure Post-Mortem Template and Review of One Incident
Review an automation incident and prevent repeats.
-
Human-in-the-Loop Design Audit: Where Approvals Add Safety and Where They Add Noise
Check approval steps for usefulness.
-
Agent Prompt Audit: Clarity, Conflicts and Missing Rules
Review an agent's instructions for ambiguity and gaps.
-
Prompt Injection Exposure Audit for Your Own Assistant (Defensive)
Identify where untrusted text can reach your assistant's instructions.
-
Agent Logging and Privacy Audit
Check what an agent logs and who can see it.
-
Workflow Automation Inventory Audit: What Runs, Who Owns It, What Breaks
List and review every automation in use.
Prompts
-
Agent Prompt Audit: Clarity, Conflicts and Missing Rules
Review an agent's instructions for ambiguity and gaps.
-
Agent Run Transcript Review: What Went Wrong and Why
Analyse a failed or odd agent run step by step.
-
Tool Description Audit: Overlap, Ambiguity and Missing Error Cases
Check an agent's tool definitions for clarity.
-
Agent Permission and Risk Review for a Proposed Automation
Assess risks before an agent is given access.
-
Automation Failure Post-Mortem Template and Review of One Incident
Review an automation incident and prevent repeats.
-
Customer-Facing Assistant Conversation Sample Audit
Review sample conversations for accuracy, tone and escalation.
-
Workflow Automation Inventory Audit: What Runs, Who Owns It, What Breaks
List and review every automation in use.
-
Data Used in Prompts Audit: Is Anything Sensitive Being Sent?
Check what information flows into prompts and models.
-
Test Suite Coverage Audit for an Agent or Assistant
Check that your tests cover the ways it can fail.
-
Prompt Change Impact Audit: Before and After Comparison
Compare two prompt versions on the same inputs.
-
Agent Quality Metrics Dashboard Audit: What to Measure and What to Ignore
Choose metrics that show whether the assistant helps.
-
Prompt Injection Exposure Audit for Your Own Assistant (Defensive)
Identify where untrusted text can reach your assistant's instructions.
-
Agent Logging and Privacy Audit
Check what an agent logs and who can see it.
-
Human-in-the-Loop Design Audit: Where Approvals Add Safety and Where They Add Noise
Check approval steps for usefulness.
No prompts match those filters.