Chrome Web Store Verified domain

Precog

ed5d4df5-15b5-52e2-ae5a-0cc7884a226c | Store v0.2.0.17
43/ 100
MEDIUM risk
+8 since v0.2.0.13
AI review verdict
Review before use

The AI review rates the findings as likely false positive, but the risk score (43/100) still counts them. The review covers version 0.2.0.8, not the analysed version 0.2.0.17.

Analysis record

Analysed
Today 22 November 2026
Version
v0.2.0.17
Artifact
SHA256 e5ae15d0…f71aea
Detail
Full list

Evidence ledger

Ranked by severity · findings with a source location link to the code viewer

3 findings

Publisher Evidence

Limited evidence

dev--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.

46
Publisher domain
mornin.org
Observed
Store verification signal
Limited signal
Limited
Extension portfolio
5
Portfolio

11 evidence rows available.

Finding Categories

3
Network

Requested Permissions

6 permissions
nativeMessaging

Exchange messages with programs outside the browser

Dangerous
activeTab
Medium
tabs
Medium
contextMenus
Low
scripting
Low
storage
Low

AI Security Report

AI Security Review

Evidence context: threat category none; evidence quality strong.

Precog is an open-source Chrome extension that extracts page context and sends prompts to Claude.ai. The evidence bundle contains 44 total findings, but none indicate malicious intent.

IoC Findings Are Garbage: The 28 IoC findings are classic XIOC extractor false positives. Domains like e.target, block.id, backdrop.click, el.dataset.id, b.modes.map, blocks.map, modal.id, and data.channel are JavaScript property access chains, not network domains. The finding XIOC-IP-0.2.0.8 is the extension's version number misread as an IP address. The only legitimate network finding is https://claude.ai/*, which directly matches the extension's stated purpose of running prompts in Claude.

Manifest Permissions Are Justified: The tabs permission (from MANIFEST-SENSITIVE-PERM-TABS in manifest.json) is legitimately required for an extension that extracts page context. Without this permission, the extension could not fulfill its documented functionality.

No Malware Indicators: There are zero malware signatures, zero obfuscation findings, and zero network findings beyond the expected Claude.ai domain. The 15 code-smell findings are all low severity and classified as noise per CVEQ's documented false positive patterns.

Counterargument: A skeptic might argue that the tabs permission is sensitive and could enable data exfiltration or tracking. However, this permission alone does not constitute malicious behavior. The extension's developer attribution ([email removed]), open-source nature, and lack of obfuscation or malware signatures indicate legitimate development. The only external domain in the IoCs (claude.ai) aligns with the extension's documented purpose. Without evidence of data transmission to suspicious domains, credential access patterns, or deceptive naming, there is no basis to classify this as malicious.

Conclusion: This extension exhibits all the hallmarks of a false positive driven by XIOC extractor noise and code-smell rules that fire on legitimate JavaScript. The findings do not reveal any actual malicious behavior.

Key Reasons

  • All IoC findings are XIOC extractor garbage (property access chains, version number)
  • Zero malware signatures and zero obfuscation findings
  • Only legitimate domain (claude.ai) matches stated extension purpose
  • Tabs permission is required for documented functionality
  • Developer attribution present with open-source claim

False Positive Considerations

  • XIOC property access chains misread as domains (e.target, block.id, etc.)
  • Version number 0.2.0.8 misread as IP address
  • Low-severity code-smell findings classified as noise
  • Tabs permission legitimately required for page context extraction

Reviewed 2026-04-27 · version 0.2.0.8; recommended action: suppress false positive; model confidence 85%.

Chrome version history

Risk trend by version

4 analyzed versions. Each point is the latest successful scan for that version; failed zero-score scans are hidden. Dates: First seen by Risky Plugins.

Selected
43
Change since first
+8
Change from previous
+8
Versions:
First analyzed version
0.2.0.8
First seen Mar 24, 2026
Risk range
35 to 43
Across analyzed versions
Latest analyzed version
0.2.0.17
First seen Nov 22, 2026
Selected version
medium
Version
v0.2.0.17
First seen Today
Risk score
43
Findings
3
Change vs previous
+8

Pick any point on the chart, then open the Source tab to read that version's code.

Source Code Viewer

Source Code Not Available

Source code is not available for this version of the extension.

About This Extension

Precog extracts structured context from the page you're on and assembles a Claude prompt from modular instruction blocks. Press Cmd+Shift+P (Mac) or Ctrl+Shift+P (Windows/Linux) on any webpage to open a command palette. Precog reads the current page and presents a set of instruction blocks: summarize, identify TODOs, draft a reply, create an Asana task, do deep research, or search for related context across Gmail, Google Drive, Slack, and other connected services. Blocks can be combined. Precog assembles the page context and selected instructions into a prompt and sends it to Claude. Example: open Precog on a Gmail thread and select the following blocks: - Search for deep context: Claude searches Gmail, Google Drive, and Slack for related threads, documents, and messages, so the response is informed by the full history rather than just the current email. - Identify TODOs: extracts TODOs from the thread. In "advise" mode, it also suggests approaches with confidence levels. - Draft a reply: generates a response based on the thread content and any context found in the search. - Create Asana task: creates a task to track follow-up, with a title, description, and link back to the email. Precog has specific integrations for Gmail, Asana, and Slack. On Gmail it reads the full thread with subjects, senders, and dates. On Asana it pulls the task title and description with formatting and links preserved. On Slack it gets the message text, sender, channel, timestamp, and permalink. On any other page, Precog captures the page title, URL, and any selected text or page content. Precog also adds a few shortcuts. Select text that contains links, right-click, and choose to open all of them in new tabs, in separate new windows, or as tabs in one new window. On Gmail, press Ctrl+Cmd+C (Mac) or Alt+Shift+C (Windows/Linux) to copy a link to the current thread to paste into Claude. Right-click a page or the Precog icon and choose "Copy URLs of all tabs in this window", or press Ctrl+Cmd+T (Mac) or Alt+Shift+T (Windows/Linux), to copy one URL per line; Gmail tabs with a thread open are copied as that same Claude-ready link. For best results, enable connectors in Claude for Gmail, Google Drive, Slack, Asana, and any other services relevant to the context being processed. Precog runs entirely in the browser. There is no server, no analytics, and no tracking. Data is only sent to Claude when a prompt is explicitly submitted. Open source (Apache 2.0): https://precog.cc

Frequently Asked Questions