crashlytics
Pass
Audited by Gen Agent Trust Hub on Mar 7, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The configuration in
.mcp.jsonusesnpxto runfirebase-tools, which is an official package from a well-known service (Google). This is a standard dependency for fetching crash data from Firebase. - [COMMAND_EXECUTION]: The skill utilizes
git logandgit blameto correlate crash locations with repository history and identify appropriate developers for assignment. These are routine development operations. - [PROMPT_INJECTION]: The skill processes external data from Crashlytics, which creates a surface for indirect prompt injection.
- Ingestion points: Crash telemetry is retrieved from the Firebase MCP server as described in
commands/crash-report.md. - Boundary markers: No specific delimiters are defined to isolate external error messages from the agent's instructions.
- Capability inventory: The skill can perform file system writes, execute git commands, and initiate subagent tasks.
- Sanitization: There is no explicit mention of filtering or sanitizing the input data from the crash reports.
Audit Metadata