Free tools Windows power users keep installed
One-click scans. No signup required.
Audit an agent’s effective read set before it indexes or searches a workspace by checking six things: the intended data scope, every discovery route, platform exclusions, the deployed identity’s permissions, positive and negative retrieval tests, and the logs and revocation controls. An index setting is only one boundary. An agent may still reach files through text search, direct reads, connector calls, directory metadata, or content already present in its conversation context.
The procedure below is platform-aware, with specific checks for Visual Studio Code workspaces, Microsoft agent deployments, ChatGPT Workspace Agents, Microsoft 365 administration, and AWS-style agent governance.
As an Amazon Associate I earn from qualifying purchases.
What a workspace read-set audit must prove
A useful audit answers a narrow question: Given this task, this deployed identity, and these enabled tools, which data can the agent actually retrieve? That is different from asking which files appear in an index. Retrieval can occur through several paths, and each path can have a different boundary.
- Resource boundary: the workspace, collection, repositories, sites, folders, or connected systems the task is allowed to use.
- Discovery surface: semantic indexes, text search or grep, file and directory listings, symbols, selected editor text, conversation history, previous tool results, direct file reads, and connectors.
- Identity: the account, service principal, application identity, or signed-in user under which each call runs.
- Action: read, search, summarize, export, modify, or perform an external operation.
- Traceability: logs that identify the agent, resource, action, and request that produced a result.
Your audit is complete only when approved examples are retrievable through the routes the agent needs and representative restricted examples fail through every route that is enabled.
#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.
1. Write the intended scope before touching settings
Start with a short access statement that another administrator could verify. Name the task, the workspace or collection, approved repositories and sites, expected file classes, and the identity that should perform retrieval. Record the owner or sponsor and the person who approves access. Also list data that must remain outside the task.
| Scope item | What to record | Why it matters |
|---|---|---|
| Purpose | One task, such as summarizing approved project documentation | Prevents broad “general research” access from becoming the default |
| Resources | Named workspace, collection, repositories, sites, and folders | Creates a testable resource boundary |
| Data classes | Approved documents, source files, issues, or tickets | Defines what a successful retrieval looks like |
| Excluded classes | Credentials, secrets, personal data, unrelated logs, generated artifacts, and large datasets | Provides negative-test targets |
| Tools and actions | Semantic search, grep, direct read, connector, export, or write operation | Reveals alternate retrieval paths |
| Identity and owner | Agent identity, sponsor, approver, and operating environment | Lets you review effective permissions rather than assumptions |
Use read-only access for a summarization task unless a separate, documented operation genuinely needs to write. Keep cross-tenant and guest paths disabled unless they are explicitly approved.
2. Map every way the agent can discover content
Do not begin with the index. Draw a route map of the running deployment and mark which routes are enabled.
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 reinstallSemantic indexing
A semantic index can retrieve relevant passages without a literal keyword match. Check which directories and file types are indexed, when the index was built, and which identity can query it. An index exclusion should be treated as a platform behavior to verify, not as proof that the underlying file is unreadable.
Text search, grep, and file search
Keyword search may expose a matching line as conversation context even when the agent never opens the full file. Test search separately from semantic retrieval. Include filename and directory searches, because metadata can disclose sensitive project names or locations.
Rank #2
- 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
Direct reads and directory structure
Agents can sometimes request a known path, list a directory, or follow a path learned from another result. A file that is absent from an index may still be readable by an explicit file-read tool. Test both an exact path and a directory listing.
Editor and conversation context
In Visual Studio Code, selected or visible editor text, symbols and definitions, conversation history, and previous tool results can become context. A user who pastes a secret into chat has already bypassed workspace exclusion settings, so include copied or previously returned content in your threat model.
Connected data sources
Connectors may query systems outside the local workspace. Identify each connector, its resource scope, and the identity used for its calls. An action constraint can limit how an action is invoked without filtering the data that the connector returns; treat returned content as subject to the connector’s effective permissions.
3. Inspect Visual Studio Code workspace boundaries
Visual Studio Code agent documentation describes several context sources: indexable workspace files, directory structure, symbols and definitions, selected or visible text, conversation history, and previous tool results. Relevant text files in the current project can be semantically indexed, while common binary files such as images and PDFs are not. When a semantic index is unavailable, agents may fall back to text search, grep, file search, and language intelligence.
Inspect the three exclusion mechanisms independently. Their effects are not interchangeable.
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
| Setting | Explorer visibility | Text search and grep | Semantic indexing |
|---|---|---|---|
.gitignore |
Not the primary Explorer-hiding control | Excluded | Excluded |
files.exclude |
Hidden | Excluded | Excluded |
search.exclude |
Visible | Excluded | Not stated as an index exclusion; verify the running platform |
Because these controls affect different surfaces, review the repository’s .gitignore, the workspace or user files.exclude rules, and search.exclude rules as separate evidence. Then test the actual agent. A hidden Explorer item is not automatically protected from a direct read, and an index exclusion does not establish that a connector or another tool cannot retrieve the same data.
Quick wins for a faster PC:
Clear out junk files and repair common Windows errorsFree Scan →Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →VS Code test set
- Create or identify one representative approved text file and one representative restricted file in each relevant directory.
- Run a semantic query for a distinctive phrase from each file.
- Run text search and grep for the same phrase, including a filename-only search.
- Ask the agent to read each exact path and to list the containing directory.
- Check whether the restricted file’s name, matching line, or content appears in the conversation or tool results.
Use harmless marker text for testing; never plant a real credential to prove that a control works.
4. Verify the identity that performs retrieval
Review permissions using the identity the deployed agent actually uses, not only the builder’s account. A local test performed while you are signed in can produce a false sense of safety when production runs as an application identity or service account.
Dedicated agent identity versus delegated user access
A dedicated identity gives you a stable subject for least-privilege grants, ownership, logging, and revocation. User-delegated access runs in the signed-in user’s context and may change when the user changes roles or belongs to new groups. In Microsoft environments, application permissions can run without an active user and may reach organizational data; delegated permissions act with the signed-in user’s permissions. Record which model applies to every retrieval route.
Effective permission review
Inspect direct grants, group or role membership, inherited permissions, tool-level scopes, downstream system permissions, and any cross-tenant or guest relationship. Review the aggregate result: an agent with several individually narrow tools can still have broad combined reach. Deny tools that are not required for the task, especially write, export, external-sharing, and high-impact actions.
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.
For a workspace summary, use a task-based, read-only role scoped to the workspace or collection and restricted to approved repositories and sites. Microsoft’s guidance states: “Grant a task-based (read-only in this case) role scoped to the workspace or collection (resource boundary) and restrict to approved repositories and sites.”
5. Run the six-step audit procedure
- Freeze the intended scope. Save the purpose, approved resources, excluded data classes, tools, identity, owner, and approver. Treat this record as the baseline for the test.
- Inspect discovery configuration. Review index coverage, ignore files, search exclusions, connector scopes, file-read tools, and any fallback search behavior. Note the date and deployment version.
- Check effective permissions. Sign in or assume the same identity used in production and inspect permissions across the workspace, repositories, sites, connectors, and downstream systems. Do not substitute the developer’s account.
- Test approved retrievals. Use representative, non-sensitive files. Confirm that semantic search, text search, direct reads, directory discovery, and required connector calls return only the approved material and expected metadata.
- Test denied retrievals. Use representative restricted files, directories, and connector records. Try the same routes, including a known path, a distinctive keyword, a filename, and a connector query. Record whether the agent receives content, a filename, a matching line, an error, or no result.
- Inspect logs and revoke. Verify that activity records identify the agent, identity, resource, action, and correlation or request ID. Disable the agent, rotate its credentials, invalidate active tokens, and remove a temporary permission; confirm that subsequent retrievals fail.
Keep the test output with the scope record. A result such as “not indexed” is insufficient unless direct read, search, connector, and metadata paths have also been tested.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.6. Apply platform-specific governance checks
Microsoft agent deployments
Microsoft recommends a named owner or sponsor and approver, documented purpose and approved data, review of aggregate and effective permissions across roles, tools, and downstream systems, detailed activity logging, and revocation tests. For workspace summarization, keep the role read-only, restrict it to approved repositories and sites, enforce data boundaries, and perform authorization checks on every retrieval call rather than only when a session starts.
ChatGPT Workspace Agents
Review the agent’s configured apps and connectors, audience, and connection owner. Publishing an agent with personal connections can allow other users to access data or perform actions through the creator’s connection. Use least privilege, limit the audience, avoid sensitive or high-impact connectors unless essential, and audit the published configuration. Connector Action Constraints can narrow an action’s use, but they do not filter or restrict the data returned by that connector, so the connector’s own authorization remains decisive.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
Microsoft 365 administration
Use the administration center’s agent-permission and activity-related security views to inspect the deployed configuration. Confirm whether each permission is application-based or delegated, then review the corresponding effective scope and activity under the identity that actually runs the agent.
Best Value
- Security Key : Protect your online accounts against unauthorized access by using FIDO2 and U2F authentication with T110. It's the world's most protective security key that works with windows, Mac OS, Linux as well as Chrome, Firefox, Edge and many other major browsers.
- Certified with the new FIDO2 standard, T110 provides the benefit of fast login and strong protection against phishing, account takeover as well as many other online attactks.
- Works with : Bank of America, Github, Google, Microsoft, DUO, Twitter, Facebook, Dropbox, Apple, ebay, BINANCE, mor and more.
- Fits USB-A port : Insert the T110 security key into the USB-A port of each service and log in conveniently with one touch
- For the driver download and user guide, please visit TrustKey Solutions Home support page.
AWS environments
AWS’s Agentic AI Lens emphasizes strong authentication for agent identities, separation between agent and human permissions, least privilege with dynamic boundaries, and regular permission audits and access reviews. Include those controls in the same scope, test, logging, and revocation record.
7. Make revocation and re-review part of the design
An audit is not a one-time approval. Repeat it whenever the workflow, tools, data scope, repository layout, connector, identity, or deployment environment changes materially. Also re-run it after a permission-group change, migration, index rebuild, or connector update.
Your revocation drill should answer four questions:
- Does disabling the agent stop new retrievals?
- Do credential rotation and token invalidation stop existing sessions or cached calls?
- Does removing a repository or group permission immediately block the affected route?
- Can you identify and remove stale grants, personal connections, guest paths, and unused tools?
Retain the prior scope, test cases, permission snapshot, logs, and revocation result so a reviewer can see what changed between audits. If a platform cannot provide a reliable activity trail or cannot enforce authorization on each retrieval, narrow the data boundary or do not connect that source.
Quick Recap
Common audit failures and fixes
| Failure | Why it happens | Fix |
|---|---|---|
| Checking only the semantic index | Fallback search or direct reads remain enabled | Test semantic, text, grep, direct path, directory, and connector routes separately |
| Testing as the builder | Production uses an application identity or another user | Repeat the test with the deployed identity and inspect effective permissions |
| Assuming an exclusion hides everything | Ignore rules affect different surfaces | Review each exclusion control and verify alternate retrieval paths |
| Trusting an action constraint | The connector still returns data outside the intended record set | Enforce scope in the connector or source authorization and test returned data |
| Allowing broad inherited access | Group, role, guest, or cross-tenant grants expand reach | Remove unreviewed paths and use a task-based, read-only boundary |
| No revocation test | Permissions are approved but never proven removable | Disable, rotate, invalidate, and remove access, then repeat a denied retrieval |
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.
