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 minuteWindows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstalliTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more
Contain the agent first, then revoke every active session and token it could use, rotate every secret it could read, remove persistence, and verify in logs that the old credentials fail. Treat an unauthorized AI agent like a compromised automation identity: disabling its process alone is not enough if refresh tokens, cloud sessions, API keys, source-control tokens, or copied secrets remain valid.
This guide gives a vendor-aware recovery sequence for Microsoft Entra ID, AWS, GitHub, endpoint workloads, and connected services. Adapt the commands and consoles to your tenant and incident policy, and preserve evidence before deleting artifacts.
What “revoke access” must accomplish
An AI agent can authenticate through several layers at once. It may have a running process, a user session, an OAuth grant, a temporary cloud role, a personal access token, an SSH key, or secrets injected into an environment variable. Revoking only one layer leaves another path open.
The Tool Desk
Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →Successful recovery has four outcomes:
- The unauthorized process and any host or workload it controls are contained.
- Existing sessions, refresh tokens, OAuth grants, temporary credentials, API keys, and source-control tokens no longer work.
- Secrets the agent could read have been replaced with distinct, least-privilege credentials.
- Audit evidence shows that old credentials are rejected and no persistence or replacement access path remains.
Do not assume that a deleted agent account invalidates application-issued cookies, cached credentials, or tokens minted for a different identity. Revoke each trust relationship separately.
#1 Best Overall
- POWERFUL SECURITY KEY: The YubiKey 5C NFC is the most versatile physical passkey, protecting your digital life from phishing attacks. It ensures only you can access your accounts
- WORKS WITH 1000+ ACCOUNTS: Compatible with popular accounts like Google, Microsoft, and Apple. A single YubiKey 5C NFC secures 100+ of your favorite accounts, including email, password managers, and more
- FAST & CONVENIENT LOGIN: Plug in your YubiKey 5C NFC via USB and tap it, or tap it against your phone (NFC), to authenticate. No batteries, no internet connection, and no extra fees required
- MOST SECURE PASSKEY: Supports FIDO2/WebAuthn, FIDO U2F, Yubico OTP, OATH-TOTP/HOTP, Smart card (PIV), and OpenPGP. That means it’s versatile, working almost anywhere you need it
- PRIMARY & SPARE KEYS: Just like having a spare house key, we recommend buying two YubiKeys - one for daily use and one as a spare. That way you’ll never get locked out of your accounts
1. Contain the agent and preserve evidence
Disable execution and isolate connected systems
- Disable the unauthorized agent, scheduler, workflow, integration, or API client. Stop its workers and prevent automatic restart.
- Isolate affected hosts, containers, virtual machines, and cloud workloads from sensitive networks while maintaining a controlled evidence channel.
- Suspend identities used by the agent, including human accounts, service accounts, workload identities, and delegated roles.
- Block suspicious egress destinations and disable newly discovered integrations until their ownership is established.
AWS guidance for agents recommends revoking affected identities and avoiding shared static keys. If a shared key was used, assume every workload that could access it is exposed until you complete the inventory.
Capture evidence before cleanup
Preserve prompts, model responses, tool-call records, agent configuration, environment-variable references, process arguments, container images, filesystem timestamps, and relevant identity, cloud, source-control, and endpoint logs. Record the time of containment, the identities involved, and the exact revocation actions. Export logs to a protected location so an attacker or compromised workload cannot rewrite the only copy.
Do not rotate or delete evidence blindly. First determine which credentials were present, where they were stored, which resources they accessed, and whether the agent created additional identities or scheduled tasks.
Free tools Windows power users keep installed
One-click scans. No signup required.
2. Revoke identity-provider sessions
Microsoft Entra ID sequence
For an affected Entra user or workload, disable the account, revoke refresh tokens, disable registered devices when the device itself is not trusted, and revoke active sessions. A password reset alone does not necessarily remove every already-issued session.
- Disable the compromised user or service identity in the Entra admin center.
- Revoke the user’s refresh tokens and sign-in sessions.
- Review registered devices and disable or remove devices that the agent could control.
- Inspect enterprise applications, app registrations, delegated permissions, application permissions, and consent grants created or modified during the incident.
- Remove unauthorized grants and rotate any client secrets or certificates associated with affected app registrations.
Microsoft warns that application-issued session cookies may require application-side revocation and can remain valid until their normal expiry. In practice, invalidate sessions in the application that issued them, clear server-side session stores, and force reauthentication where the application supports it.
Rank #2
- 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.
Check non-Entra identity paths
Repeat the same inventory for other identity providers, single sign-on systems, local accounts, device-management platforms, and secrets managers. Look for active sessions, remembered devices, recovery methods, API clients, and OAuth grants. Revoke the grant rather than merely changing a display name or disabling the visible agent.
3. Revoke cloud and source-control credentials
AWS temporary sessions and roles
AWS provides immediate revocation for temporary delegation sessions; the action is recorded in CloudTrail. Revoke sessions issued to the affected role, user, or delegated agent, then review CloudTrail for use after the suspected compromise.
AWS IAM can invalidate temporary role credentials issued before a selected time. Set that cutoff after the last known legitimate activity, understanding that every session issued before the cutoff will require reauthentication. This is useful when you cannot enumerate every temporary session individually.
- Identify the IAM principal, assumed-role session names, and source workload used by the agent.
- Revoke temporary delegation sessions and set the IAM temporary-credential validity cutoff where appropriate.
- Deactivate or delete exposed access keys only after you have identified legitimate dependencies and prepared replacements.
- Review inline and managed policies, trust policies, permission boundaries, role chaining, and newly created users or roles.
- Use CloudTrail to confirm the revocation event and investigate any calls made after containment.
GitHub tokens and app access
GitHub’s revocation endpoint covers classic and fine-grained personal access tokens, OAuth-app tokens, GitHub-App tokens, and refresh tokens. Revoke every token associated with the affected user, application, installation, or workflow, not just the token visible in the agent configuration. A revoked credential must be replaced before automation can resume.
Also inspect repository deploy keys, organization OAuth approvals, GitHub App installations, Actions secrets, runner registrations, webhooks, branch-protection changes, and workflow files. A token may be gone while a modified workflow or deploy key continues to provide access.
Rank #3
- POWERFUL SECURITY KEY: The YubiKey 5 NFC is the most versatile physical passkey, protecting your digital life from phishing attacks. It ensures only you can access your accounts
- WORKS WITH 1000+ ACCOUNTS: Compatible with popular accounts like Google, Microsoft, and Apple. A single YubiKey 5 NFC secures 100+ of your favorite accounts, including email, password managers, and more
- FAST & CONVENIENT LOGIN: Plug in your YubiKey 5 NFC via USB and tap it, or tap it against your phone (NFC), to authenticate. No batteries, no internet connection, and no extra fees required
- MOST SECURE PASSKEY: Supports FIDO2/WebAuthn, FIDO U2F, Yubico OTP, OATH-TOTP/HOTP, Smart card (PIV), and OpenPGP. That means it’s versatile, working almost anywhere you need it
- PRIMARY & SPARE KEYS: Just like having a spare house key, we recommend buying two YubiKeys - one for daily use and one as a spare. That way you’ll never get locked out of your accounts
Other cloud and SaaS credentials
Apply the same method to every connected provider: enumerate the principal, revoke active sessions and grants, invalidate temporary credentials, rotate long-lived keys, and inspect audit logs. Include data warehouses, ticketing systems, messaging platforms, package registries, monitoring tools, and internal APIs if the agent could reach them.
4. Rotate every secret the agent could read
Rotation must be based on reachability, not on the files you happened to find. If the agent could read a host, container, repository, build log, secret store, or environment variable, include the secrets exposed through that path.
Secret inventory
- User and service-account passwords.
- Cloud access keys, role credentials, and instance or workload identity bindings.
- API keys, OAuth client secrets, refresh tokens, and webhook signing secrets.
- SSH private keys, repository deploy keys, signing keys, and package-publishing credentials.
- Database usernames and passwords, connection strings, and encryption-key references.
- CI/CD variables, build-system tokens, runner registration tokens, and deployment credentials.
- Certificates and private keys used for mutual TLS or application signing.
- Recovery codes, backup credentials, and secrets embedded in scripts, notebooks, images, or logs.
Use a safe replacement pattern
- Create a separate identity for each agent, workload, or deployment rather than reusing a team-wide credential.
- Grant only the actions and resources required for that function. Remove administrative, wildcard, and cross-environment permissions.
- Prefer short-lived credentials issued just before use and expired immediately afterward.
- Store secrets in a managed secret store where available; do not put replacements in source code, prompts, chat transcripts, images, or shared environment files.
- Deploy the replacement to the approved workload, test the minimum required operation, and then disable the old credential.
- Update dependent services, workers, scheduled jobs, and rollback plans so an old secret is not silently reintroduced.
AWS specifically cautions against shared static keys for agent authentication. Distinct identities reduce blast radius and make audit events attributable to one workload.
Rotation order that avoids a second outage
For a multi-service system, rotate the highest-risk administrative and signing credentials first, then data-plane credentials, then lower-risk integrations. Keep old and new values available only during a short, controlled overlap when a dependency cannot switch atomically. Once the replacement is confirmed, revoke the old value and search logs and repositories for its fingerprint or last four characters.
5. Remove persistence and unauthorized changes
An agent that can log back in has not been contained. Search for persistence in both control planes and hosts.
Quick wins for a faster PC:
Repair Windows errors before they cause bigger problemsFix Now →Scan for outdated or missing drivers - takes under a minuteDriver Scan →Rank #4
- 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.
Identity and cloud persistence
- New users, service principals, roles, policies, permission-boundary changes, and trust-policy changes.
- New access keys, certificates, OAuth grants, app installations, and delegated permissions.
- Modified security groups, network routes, firewall rules, storage policies, and cross-account trust.
- New scheduled functions, event rules, automation documents, or startup hooks.
Endpoint and workload persistence
- Scheduled tasks, cron entries, systemd services, launch agents, startup scripts, and container restart policies.
- Modified shell profiles, package hooks, browser extensions, IDE tasks, notebooks, and development-tool configuration.
- New binaries, scripts, model tools, plugins, webhooks, reverse tunnels, and command aliases.
- Changed CI/CD workflows, build definitions, runner labels, deployment manifests, and image entrypoints.
Compare current configuration with a known-good baseline. Preserve suspicious files for investigation before removing them, and document who approved each deletion.
6. Verify that revocation worked
Use logs as a controlled test
Review identity-provider, cloud, source-control, application, and endpoint logs for the period beginning before the first suspicious activity and extending beyond revocation. Confirm the source IPs, user agents, device IDs, role-session names, repositories, API methods, and resources touched.
Where policy permits, perform a controlled test using a revoked credential or a safe canary account. The expected result is an authentication or authorization failure. Test the replacement credential separately and confirm it succeeds only for the approved principal, resource, network, and operation.
AWS documents that temporary-access revocation is logged in CloudTrail. Use that event, subsequent denied calls, and the absence of unexpected successful calls as evidence—not as proof that no undiscovered path exists.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
Verification checklist
- Old refresh tokens, sessions, OAuth grants, temporary credentials, API keys, and source-control tokens are rejected.
- Application session cookies are invalidated or have expired under the application’s revocation policy.
- Replacement credentials are unique, least-privilege, and stored in the approved secret system.
- No unauthorized user, role, key, device, webhook, workflow, scheduled task, or startup item remains.
- Logs show no successful access from the old principal after the cutoff.
- Backups, images, caches, build artifacts, and logs do not retain usable copies of rotated secrets.
- Owners can explain the recovery path if the new credential fails.
7. Harden administrator access after recovery
Require phishing-resistant FIDO2 MFA for administrators and other high-impact identities. Microsoft states: “FIDO2 security keys provide hardware-backed, phishing-resistant authentication that protects against credential theft and unauthorized access.” Enroll compatible security keys, keep a secure recovery method, and test account recovery before closing the incident.
Best Value
- FIDO2 & Passkey Ready: Business-ready and FIDO2 L1 certified. This key is supported by major management suites and is ideal for both individual and enterprise deployment. Works seamlessly with Gmail, Facebook, GitHub, Dropbox, Coinbase, and more.
- Universal Connectivity (USB-A ): Features a built-in USB-A connector—simply unfold the key and plug it into your compatible PC or laptop for seamless authentication on the go.
- Dedicated Manager App: Use the Thetis Manager App for the initial hardware PIN setup. Setting the PIN on the device first ensures a smooth registration process. Once the PIN is configured, you can begin registering the key across your favorite FIDO2-compatible online services.
- Ultra-Durable & Portable: Featuring a rotating metal cover, this key is water, crush, and tamper-resistant. It fits easily on a keychain and requires no batteries or network connectivity.
- Check FIDO2 compatibility before purchase - Known limitations: ID Austria is not supported (requires FIDO2 Level 2). Windows Hello login only works with Windows Enterprise editions that support Entra ID, and NFC is NOT supported.
Use separate administrator accounts, remove standing privilege where possible, require approval for elevation, and alert on new credentials, consent grants, role changes, and unusual agent behavior. Give each agent a documented owner, purpose, identity, permission set, credential lifetime, and shutdown procedure.
Control comparison: what each measure stops
| Control | Revocation scope | Time to invalidate | Credential lifetime | Least-privilege support | Auditability | Blast radius | Distinct agent identity |
|---|---|---|---|---|---|---|---|
| Disable agent process | One process or workload | Immediate if it cannot restart | Does not change existing tokens | Depends on its identity | Endpoint and scheduler logs | Low for the process; high if persistence remains | Not stated |
| Disable user or service identity | Sign-ins for that identity | Usually immediate for new sign-ins | Existing sessions may remain | Policy-dependent | Identity-provider logs | Limited to the identity, unless credentials were shared | Yes, if the identity is unique |
| Revoke refresh tokens and sessions | Token-based sessions covered by the provider | Provider-dependent; application cookies may persist | Reauthentication required after invalidation | Not a permission control | Identity and application logs | Session scope | Not by itself |
| Revoke temporary cloud credentials | Delegated and temporary role sessions | Immediate revocation or selected cutoff | Short-lived by design | IAM policy supports it | Cloud audit logs; AWS records revocation in CloudTrail | Role and session scope | Yes when each workload assumes its own role |
| Rotate long-lived secrets | Every system using the secret | After dependent services switch and old value is disabled | Can be short-lived if redesigned | Scopes can be reduced on replacement | Secret-manager and service logs | Potentially broad when a secret is shared | Yes when replacements are per workload |
| FIDO2 MFA | Administrator authentication | At next authentication challenge | Hardware-backed key remains until removed | Works with separate privileged roles | Identity-provider and device-registration logs | Protects privileged entry points | Applies to the administrator, not the agent |
Common failure modes and fixes
Stopping the model but leaving tokens active
Symptom: The agent is offline, but API calls continue from its host or a second worker.
Fix: Revoke sessions, refresh tokens, OAuth grants, temporary credentials, API keys, and source-control tokens, then inspect logs for alternate principals.
Changing one password and declaring recovery
Symptom: A password reset is complete, yet an application session or app-issued cookie still works.
Fix: Revoke provider sessions and perform application-side session invalidation; remove remembered devices and grants that can mint new sessions.
Rotating only the secret found in configuration
Symptom: A later discovery shows the agent had read database credentials, SSH keys, or CI variables elsewhere.
Fix: Reconstruct the agent’s read permissions and rotate every reachable secret, including copies in images, caches, logs, and backups.
Reusing a shared replacement key
Symptom: You cannot attribute activity or revoke one workload without breaking others.
Fix: Issue distinct, short-lived, least-privilege credentials per agent or workload and document ownership.
Deleting persistence before collecting evidence
Symptom: The visible task is gone, but the incident timeline and original payload cannot be reconstructed.
Fix: Preserve prompts, tool calls, configurations, timestamps, and logs first; then remove persistence under an approved response plan.
Skipping the negative test
Symptom: The team assumes revocation succeeded because no one has tried the old credential.
Fix: Conduct a safe, controlled rejection test and verify denied events in the relevant logs.
Do these 3 things before closing this tab:
1Scan for outdated or missing drivers - takes under a minute2Clear out junk files and repair common Windows errors3Fix the driver behind crashes, sound loss and screen glitchesQuick Recap
Incident closeout checklist
- Contain the agent, isolate affected hosts, suspend compromised identities, and preserve evidence.
- Revoke identity-provider sessions, refresh tokens, devices, OAuth grants, and application cookies where applicable.
- Invalidate cloud delegation sessions and temporary role credentials; review cloud audit events.
- Revoke source-control tokens, deploy keys, app installations, workflow secrets, and runner access.
- Inventory and rotate every reachable password, key, token, certificate, signing secret, database credential, and CI/CD variable.
- Remove unauthorized users, keys, grants, workflows, webhooks, scheduled tasks, startup items, and configuration changes.
- Test that old credentials fail and replacements work only from approved principals.
- Require phishing-resistant FIDO2 MFA for administrators and test account recovery.
- Record the incident timeline, affected resources, replacement owners, expiration dates, and monitoring changes.
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.
