Repository navigation
B86: a check before sharing - #73
Merged
Merged
Conversation
The share sheet lists email addresses and phone numbers in what a trail
file or share link would share, each ticked; unticking one leaves it out
("[email left out]") without changing the trail. Agents get an opened
trail file marked as untrusted data. The red-team set gains an injected
trail file.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011V7Bv7WZ9Q92MZoNWYwevQ
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
B86. A check before sharing, for trail files and share links.
What it checks (your choices: links and files; contact details only)
In the share sheet (your choice: warn, untick to leave out)
Trails private by default: already true. Nothing is shared unless you share it. No change.
Red-team set (B58)
test/e2e/site/redteam/injected.trail.md: a shared trail whose note, question and reply tell AI agents to send data to the attacker server.18-redteampart C: the file opens read only; MCPget_research_sessiongets it with a first line marking it untrusted; Ask about this trail with the compromised model shows the answer as text, with no attacker image loaded and zero requests to the attacker.Code
core/share-check.js(no platform imports):findContacts,snapshotTexts,sessionTexts,redact.main.js:trail:share-check,leaveOutfor files and links.core/trail.js(fromFilein the session view),core/research.js(the untrusted line).renderer/app.js,index.html,styles.css: the two lists. Tokens only.Tests
test/share-check-demo.js(npm run test:share-check, intest:unit): 6 checks.test/e2e/54-share-check.js(10 checks), in both CI e2e lists.18-redteam: 4 more checks.test:unit, and e2e 15, 18, 37, 38, 40, 44 and 53 still pass.Generated by Claude Code