Quick wins for a faster PC:
Repair Windows errors before they cause bigger problemsFix Now →Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Clear out junk files and repair common Windows errorsFree Scan →Stop an AI agent from taking unauthorized actions by enforcing permissions in the software that executes its tools—not by relying on a system prompt. Give it only the capabilities and access required for the task, check every action against an external policy, and require approval for consequential operations. Then test those controls against malicious inputs and ordinary mistakes.
Why an AI agent may take an action you did not intend
An agent can misuse its access because it misunderstands a request, makes a model error, has more capability than the task requires, or is manipulated by instructions hidden in content it reads. An email, webpage, ticket, or document can contain malicious text that attempts to redirect the agent. NIST calls this kind of indirect prompt injection agent hijacking.
As an Amazon Associate I earn from qualifying purchases.
A prompt can explain what the agent should do, but it cannot reliably enforce what the agent is allowed to do. The durable boundary is in the tool-execution path and the connected service: those systems must authorize each request, regardless of what the model proposes. OWASP puts it plainly: “Implement authorization in downstream systems rather than relying on an LLM to decide if an action is allowed or not.”
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
Build the controls in this order
1. Inventory every action the agent can reach
List its tools and connectors, the identity or credentials behind each one, the files and records it can access, network destinations, and possible external effects. Include indirect routes: a general shell, browser, or URL-fetch function can expose far more than a narrowly defined tool.
#1 Best Overall
- POWERFUL SECURITY KEY: The Security Key C NFC is the essential physical passkey for protecting your digital life from phishing attacks. It ensures only you can access your accounts.
- WORKS WITH 1000+ ACCOUNTS: Compatible with Google, Microsoft, and Apple. A single Security Key C NFC secures 100 of your favorite accounts, including email, password managers, and more.
- FAST & CONVENIENT LOGIN: Plug in your Security Key C NFC via USB-C and tap it, or tap it against your phone (NFC) to authenticate. No batteries, no internet connection, and no extra fees required.
- TRUSTED PASSKEY TECHNOLOGY: Uses the latest passkey standards (FIDO2/WebAuthn & FIDO U2F) but does not support One-Time Passwords. For complex needs, check out the YubiKey 5 Series.
- BUILT TO LAST: Made from tough, waterproof, and crush-resistant materials. Manufactured in Sweden and programmed in the USA with the highest security standards.
Remove capabilities that the intended task does not require. For example, an agent that only needs to read email should not also receive send or delete functions. OWASP identifies excessive functionality, permissions, and autonomy as sources of excessive agency; its LLM06:2025 guidance recommends reducing them.
2. Narrow tools and credentials
Prefer purpose-built operations such as “write this approved file” over unrestricted command execution. Where possible, use read-only access, scope credentials to a user and task, separate identities across environments or trust levels, and limit access to specific resources.
Apply those limits in the connected service’s own authorization system. A tool description or prompt that says “do not modify other files” is not a substitute for credentials that prevent the agent from doing so. OWASP’s AI Agent Security Cheat Sheet covers least privilege, tool authorization, and approval integrity.
Recommended Free Tools
3. Check authorization before every action executes
Put a policy check in the execution path, separate from the model’s decision-making. For each proposed call, check the authenticated actor, operation, target, parameters, risk level, and whether required approval is valid. Deny unknown actions by default, and have the downstream service validate authorization on every request as well.
Rank #2
- POWERFUL SECURITY KEY: The Security Key NFC is the essential physical passkey for protecting your digital life from phishing attacks. It ensures only you can access your accounts.
- WORKS WITH 1000+ ACCOUNTS: Compatible with Google, Microsoft, and Apple. A single Security Key NFC secures 100 of your favorite accounts, including email, password managers, and more.
- FAST & CONVENIENT LOGIN: Plug in your Security Key NFC via USB-A and tap it, or tap it against your phone (NFC) to authenticate. No batteries, no internet connection, and no extra fees required.
- TRUSTED PASSKEY TECHNOLOGY: Uses the latest passkey standards (FIDO2/WebAuthn & FIDO U2F) but does not support One-Time Passwords. For complex needs, check out the YubiKey 5 Series.
- BUILT TO LAST: Made from tough, waterproof, and crush-resistant materials. Manufactured in Sweden and programmed in the USA with the highest security standards.
A risk label does not itself authorize a tool call. The execution component still needs to confirm the actor’s permissions and any required approval for the exact operation. OWASP’s security guidance recommends binding approval to details such as the actor, tool, target, normalized parameters, timestamp, and expiry so that approval for one action cannot be reused for another.
4. Set approval thresholds outside the agent
Define in policy which operations require a person’s confirmation; do not ask the agent to decide whether it needs approval. Consider requiring review before sending messages, publishing content, making purchases or transfers, deleting records, changing permissions, or modifying production systems.
Show the reviewer the actual operation, target, and relevant parameters before they approve. The execution layer should check that the approval matches the action being attempted. If the policy check or approval system is unavailable, block the consequential operation rather than letting it proceed unchecked. Approval is one layer of authorization, not permission by itself.
5. Treat content the agent reads as untrusted data
Webpages, email, documents, tickets, and tool outputs can contain instructions written by someone other than the user or developer. Keep that content separate from trusted policy where possible. Extract only the fields the workflow needs, validate them against a schema and policy, and prevent retrieved text from adding tools, changing permissions, or choosing new recipients and destinations.
Rank #3
OpenAI’s agent safety guidance recommends structuring workflows so untrusted data does not directly drive agent behavior. Its prompt-injection guidance also describes layered protections and cautions that guidance may not prevent every attack. These measures reduce risk; they do not make prompt injection impossible.
6. Log activity, limit damage, and prepare recovery
Record who or what requested an action, the tool and parameters, target resource, policy decision, approval details, and outcome. Monitor both the agent’s tool calls and the downstream systems that carry them out; the latter can reveal effects that a tool log alone misses.
Set appropriate limits on action volume, retries, or spending. Keep a practical way to revoke credentials and disable tools quickly. Logging and rate limits can help detect or contain damage, but they do not replace authorization checks.
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 & 117. Test the boundaries continuously
Test ordinary tasks as well as adversarial cases involving malicious email, documents, webpages, compromised tools, and ambiguous requests. Check whether the agent can reach a prohibited operation, whether the policy layer blocks it, and whether approval is tied to the exact action and parameters. Repeat tests when tools, permissions, workflows, or attack methods change.
Rank #4
- Ultra-Compact FIDO2 Security Key - Plug-and-stay or carry on a keychain. This USB-A hardware security key offers portable, always-on protection for desktop and mobile use. (Item Size: 0.75 X 0.74 IN x 0.25 IN)
- USB-A Hardware Key for All Devices - Works with USB-A ports on PC, Mac, Android, and other laptop/notebook device. Enables secure, cross-platform login with FIDO2.0 passkey support.
- FIDO Certified Security Key - Meets FIDO and FIDO2 standards. Works with Google, Microsoft, GitHub, Dropbox, and more. Please check service compatibility before purchase.
- Passwordless Login with Passkey - Supports passkey login via WebAuthn and CTAP2. Enjoy password-free sign-ins where supported. Not all websites or services currently support passkeys.
- Advanced Multi-Factor Authentication - Offers 200 FIDO2 passkey slots and 50 OATH-TOTP slots. Strong, flexible 2FA/MFA support across various apps and authentication platforms.
NIST’s agent-hijacking evaluation article emphasizes task-specific and adaptive evaluations, including multiple attempts and continued testing as attacks evolve. It describes evaluations involving agents powered by Claude 3.5 Sonnet, released in October 2024; those results should not be read as a current model ranking or as a universal probability of attack success.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Choose controls according to the impact of an action
The right balance depends on what the agent can affect and how difficult an action is to reverse. Use stronger checks as impact, irreversibility, and exposure to other people increase.
| Action class | Practical control | Example |
|---|---|---|
| Read-only or analysis | Grant access only to the necessary data and destinations; log access. | Summarize a selected folder of documents. |
| Low-risk, reversible change | Use a narrowly scoped tool and validate the target and parameters in policy. | Update a draft in a designated workspace. |
| External, financial, administrative, or hard-to-reverse action | Require a person to approve the exact action; recheck authorization and approval at execution. | Send an email, transfer money, delete records, change access, or deploy to production. |
This is a risk-based design, not a universal list of safe and unsafe actions. A seemingly reversible change can still have serious consequences if it affects the wrong account, recipient, or production resource.
Free tools Windows power users keep installed
One-click scans. No signup required.
Where common safeguards fit—and where they do not
- System prompts: useful for communicating intended behavior, but not an access-control boundary.
- Input filters and model training: useful layers against unsafe or manipulative content, but do not prevent every attack or model error.
- Tool wrappers and policy gateways: useful places to validate proposed calls before execution; they must be backed by real permissions and checks in connected services.
- Human approval: important for consequential operations when it is informed, tied to the exact action, and checked at execution time; it does not grant an agent broader access by itself.
- Monitoring and rate limits: useful for detection and containment, not substitutes for preventing unauthorized actions.
Product defaults are not a general security guarantee. For example, Anthropic describes Claude Code as read-only by default in its initialized directory and says it requires approval before modifying code or systems; that is a vendor-specific behavior, not a default shared by AI agents generally. See Anthropic’s framework for developing safe and trustworthy agents.
Quick Recap
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.
