The Tool Desk
Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →An AWS audit agent can inspect the account without being authorized to change it—but “can’t touch anything” needs a precise definition. Give the agent a dedicated identity with only the read actions its audit requires, exclude write and permission-management actions, and check every tool and runtime role that could provide another route to AWS. Read-only access can still expose sensitive information, so define what the agent may see as carefully as what it may change.
What “can’t touch anything” means in AWS
AWS IAM policies determine which actions a principal may perform on which resources and under what conditions. For an audit agent, the practical goal is to authorize only the inspection actions needed to answer specific audit questions. A role with no authorized write actions is unable to make changes through that role, but this does not prove that the entire agent system has no other route to change AWS.
As an Amazon Associate I earn from qualifying purchases.
Nor does read-only mean harmless: permitted calls may reveal sensitive configuration or data. Define the agent’s visibility as well as its ability to modify resources. AWS’s IAM best practices recommend least privilege—granting only the permissions required for a task.
Crashes, No Sound, or Screen Glitches?
Random freezes, missing sound and display glitches usually trace back to one bad driver. Find and replace yours safely.Free scan · under a minutePC Slower Than It Used to Be?
A free scan shows the junk files, broken settings and background clutter dragging Windows down - then fixes them in one click.Free scan · Windows 10 & 11Design the audit role around the questions
- Write down the audit questions. For each one, identify the AWS APIs the agent must call. “Audit security” is not a permission specification; checking a particular setting in a particular service is much more actionable.
- Create a dedicated workload identity. Do not reuse a human administrator’s credentials or an application role with unrelated duties. Where the architecture allows, use temporary role credentials; AWS recommends temporary credentials for workloads.
- Build a custom allow-list. Permit the actions needed to answer the defined questions. Scope them to resource ARNs and add conditions where the relevant service supports them. Check the service authorization documentation: some actions do not support resource-level scoping and require
Resource: "*". - Keep change and permission-management actions out. Do not authorize actions that create, update, delete, or alter permissions unless they are separately required and explicitly approved. An audit agent intended to report findings should not receive remediation rights.
- Test both sides of the boundary. Verify that expected inspection calls succeed and representative mutation calls are denied. This is a recommended validation practice, not a claim that a particular agent or policy has been tested here.
- Refine from observed activity. Use CloudTrail events to see what the agent actually calls, and use IAM Access Analyzer policy generation to help identify required actions. Validate the result and remove permissions that are unused.
What AWS’s CloudTrail read-only example illustrates
AWS documents a CloudTrail-specific policy example allowing cloudtrail:Get*, cloudtrail:Describe*, cloudtrail:List*, and cloudtrail:LookupEvents with Resource: "*". AWS says that example does not grant CreateTrail, UpdateTrail, StartLogging, or StopLogging. It demonstrates that an identity can inspect CloudTrail without being allowed to create or reconfigure trails; it is not a complete policy for auditing other AWS services. See the CloudTrail identity-based policy examples.
#1 Best Overall
Do not copy the example wholesale just because it is labeled read-only. Wildcard actions and Resource: "*" can grant broader visibility than the audit needs. Start from the actual questions, then check which actions can be scoped to specific resources.
Compare the main design choices
| Choice | What it means | Trade-off |
|---|---|---|
| Dedicated custom role | Permissions are tailored to the audit’s required actions and resources. | More work to define and maintain, but easier to review against the task. |
| Broad managed read-only policy | A convenient AWS-managed permission set may cover more services or actions than the audit requires. | Faster to attach, but less tailored. AWS recommends moving toward use-case-specific least privilege; managed policies can change over time. |
| Direct AWS API tools | The agent calls AWS through its audit identity. | Fewer runtime permission paths to review. |
| Bedrock Agent action groups | Bedrock invokes tools through action groups, commonly backed by Lambda. | Adds service-role and Lambda resource-policy boundaries that must be reviewed separately. |
| Report-only recommendations | The agent describes a finding and proposes a fix; an authorized person or independent deployment pipeline applies it. | Keeps remediation outside the agent’s authorized actions. Automatic remediation changes the safety claim and needs separate authorization and testing. |
A managed policy is not a permanent contract: AWS may update its permissions. Before relying on one, inspect its current default version and review what it grants. The AWS ReadOnlyAccess managed policy reference describes the policy’s current permissions.
Check every permission path when using Bedrock Agents
Bedrock Agents are one possible runtime, not a requirement for an AWS audit agent. If you use them, the audit role is only one part of the permission picture. AWS documents a Bedrock Agent service role that may need access to the model, S3 objects containing action-group schemas, and a knowledge base; additional permissions may apply for features such as collaboration, provisioned throughput, guardrails, or encryption. An action-group Lambda also needs a resource-based policy that permits Bedrock to invoke it. See Amazon Bedrock Agent permissions.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
Review the boundaries independently rather than treating the agent’s audit identity as proof that the whole system is read-only:
Rank #3
- The role used by the agent to call AWS APIs.
- The Bedrock service role, if Bedrock Agents are used.
- The Lambda execution role and resource-based policy for each action group.
- Tool code, credential forwarding, and any cross-account role assumptions.
Keep findings separate from execution
A safer report-only design lets the agent identify a problem and explain a proposed fix without having permission to apply it. A human reviewer or a separately authorized deployment pipeline can decide whether to make the change. If the agent can execute remediations, that is a different access model—not a read-only audit—and its change permissions and safeguards must be separately authorized and validated.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Revisit the policy as the audit changes
Permissions that fit today’s audit may become excessive or incomplete as the questions, tools, or AWS services change. Review the role against observed CloudTrail activity, validate the generated policy before adopting it, and check managed-policy permissions again if the design relies on them. The aim is not to make a policy look read-only by name; it is to make the allowed actions, visible data, and remaining permission paths explicit and reviewable.
Quick Recap
Best Value
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
Quick wins for a faster PC:
Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Clear out junk files and repair common Windows errorsFree Scan →
