sentry
Pass
Audited by Gen Agent Trust Hub on Sep 4, 2026
Risk Level: SAFEDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [DATA_EXFILTRATION]: The skill reads the Sentry authentication token from the user's home directory and sends it to Sentry's API to authenticate requests. This is the intended behavior for a Sentry integration tool.
- Evidence:
lib/auth.jsreads~/.sentryclircusingreadFileSyncand extracts thetokenvalue via regular expression.lib/auth.jsincludes this token in theAuthorization: Bearerheader for allfetchrequests tohttps://sentry.io/api/0.- [INDIRECT_PROMPT_INJECTION]: The skill ingests data from an external source (Sentry) that often contains attacker-controlled or untrusted content, such as error messages, request bodies, and stack traces. If this content contains instructions for the AI agent, it could lead to indirect prompt injection.
- Ingestion points: Sentry API responses fetched in
scripts/fetch-event.js,scripts/fetch-issue.js,scripts/list-issues.js,scripts/search-events.js, andscripts/search-logs.js. - Boundary markers: Absent. The fetched content (e.g., event messages, exception values, and breadcrumbs) is formatted and displayed to the agent context without delimiters or instructions to ignore embedded commands.
- Capability inventory: The skill environment allows for network operations to the Sentry API and reading local configuration files.
- Sanitization: Absent. The scripts render retrieved metadata and message strings directly into the terminal output or JSON response for the agent to process.
Audit Metadata