Claude Cleaner
Score-based assessment (minimal risk, 0/100). An AI review is available in the AI review tab.
Analysis record
- Analysed
- 3 weeks ago 29 October 2026
- Version
- v1.1.0.12
- Artifact
- SHA256 b9e92e56…b706d0
- Detail
- Full list
No findings
The analyzer reported nothing for this version.
Publisher Evidence
Limited evidencedev--4812273f-f26b-9baa-8970-9fd77e4c9a41
Publisher identity, store signals, distribution reach, and warning signals used for context. Treat this as supporting evidence, not a clean bill of health.
Chrome does not expose the same publisher verification data as IDE stores, so this score is deliberately conservative.
11 evidence rows available.
No findings
The analyzer reported nothing for this version.
AI Security Report
AI Security Review
Evidence context: threat category none; evidence quality absent.
Claude Cleaner is a Chrome extension published under the developer contact [email removed], version 1.1.0.12, described as adding a "Delete all chats" button to claude.ai. That description is the entire factual record. There is no manifest extract, no file inventory, no content script path, and no line count to work from.
The permission data is blank. Both the permissions list and the host_permissions list come back empty. For an extension that exists to place a control inside one specific site, that is the first thing worth reading, and it is missing. An extension that only runs a content script normally keeps its origin match inside content_scripts rather than in host_permissions, so an empty host_permissions array does not prove the script is idle on every tab. It proves nobody can see whether the match is https://claude.ai/* or <all_urls>. That single string decides whether a bug in the script touches one chat page or every site where you are logged in, including mail and banking tabs.
The same hole runs through the code-level record. Every category sits at zero: no malware signature, no network finding, no obfuscation flag, no embedded secret, no code-smell hit, no indicator of compromise. A content script that clears a React-based chat interface by locating controls, looping over rows and clicking has to read DOM state and usually trips at least one low-severity rule along the way. Minified bundles almost never come back with zero extracted domain strings. So zero across the board is what a short clean script looks like, and it is also what a scan that never ingested the package looks like. With no file count in the record, those two outcomes are indistinguishable from here.
A fair skeptic would say zero findings is the best possible outcome and that a helper this narrow has nothing to hide. That holds until you weigh what the helper actually does. It automates irreversible deletion inside a logged-in account. A selector that matches more elements than the author intended, or a loop that keeps clicking after the list changes, removes chat history Claude's interface will not restore. Chat-deletion helpers are also easy to clone; a listing with a near-identical name would do the same damage with none of the original author's care. Neither failure shows up in a signature scan, which is why the empty manifest is the thing to fix here, not the finding count.
One useful negative does exist. With no network endpoints and no external domains recorded, nothing in the record shows this extension shipping chat content off the machine. A script that only manipulates the page it is injected into is well within normal extension behaviour.
What would settle it: the content_scripts match pattern, the file list with line counts, and the specific code that finds and clicks the delete control. Until those are in the record, the zero should be read as unmeasured, not as a clean result.
Key Reasons
- Zero findings in every category, including indicator, code-smell and manifest checks, with no file count or scan metadata showing the package was actually read.
- Empty permissions and host_permissions arrays leave the content_scripts match pattern for claude.ai unrecorded, so the extension's page access cannot be verified.
- The stated behaviour, automating deletion inside a logged-in account, is destructive enough that the clicking code needs direct inspection.
- No network endpoints or external domains are recorded, so nothing indicates chat content leaves the browser.
False Positive Considerations
- No scan metadata or file inventory, so an absence of findings cannot be told apart from an absence of analysis
Reviewed 2026-11-01 · version 1.1.0.12; recommended action: reanalyze; model confidence 55%.
Source Code Not Available
Source code is not available for this version of the extension.
About This Extension
Frequently Asked Questions
Similar Extensions
Related extensions from the same publisher or marketplace