Skip to content

Security: iamsonal/aiAgentStudio

Security

SECURITY.md

Security

Report a vulnerability

Do not open a public issue for a suspected vulnerability.

  1. Open the repository's Security tab.
  2. Click Report a vulnerability. This opens a private advisory that only the maintainer sees.
  3. Include the affected package and version, the steps to reproduce it, and what an attacker could do.

The maintainer aims to reply within 48 hours and to send a first assessment within one week. The time to a fix depends on how severe the problem is.

Supported versions

Security fixes target the current Pluto Core 1.0 release. Development builds from before a release are not maintained.

Run Pluto safely

The Pluto docs explain each of these:

  1. Keep model API keys and other secrets in Named Credentials and External Credentials. Never put them in instructions, records, or code. See Connect a model.
  2. Give the operator, approver, execution user, and sensitive-data permission sets only to the users who need them. See Give people access.
  3. Require approval for tools that change records or call outside systems, unless an authorization policy you reviewed allows the call. See Require approval.
  4. Check the retention settings before you send customer data to a model provider. See Retention and erasure.
  5. Turn on Core's recovery and retention schedules, and check Agent Core Health. See Check health.
  6. Limit who holds Modify All Data. The trust vault keeps real values as plain text until retention removes them. Use Salesforce Shield Platform Encryption if you need encryption at rest. See Sensitive data.
  7. Do not rely on tokens alone. Core turns emails, phone numbers, card numbers, and classified fields into tokens, but other personal data in free text reaches the model as written. Classify sensitive fields, return records as record sets, and keep each tool's Input Schema closed with "additionalProperties": false. See Sensitive data.
  8. Never change run, command, or event records by hand. Use the supported actions in Fix a stuck run. In Apex, use AgentRunService to change a run and the paged AgentTraceService methods to read its history.

There aren't any published security advisories