replit-core-workflow-b
Pass
Audited by Gen Agent Trust Hub on Sep 9, 2026
Risk Level: SAFEDATA_EXFILTRATIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [DATA_EXFILTRATION]: The skill performs network requests to official Replit domains (replit.com and replit.dev) to manage team members, audit seats, and trigger deployments. These operations align with the skill's stated administrative purpose and target a well-known service provider, which is considered safe behavior.
- [COMMAND_EXECUTION]: The skill utilizes curl commands via the Bash tool to retrieve audit logs and export member activity to CSV files. These commands are restricted to official Replit API endpoints.
- [INDIRECT_PROMPT_INJECTION]: The skill ingests external data from the Replit API (such as member lists and Repl metadata) which is then processed by the agent. This represents a standard attack surface for management tools.
- Ingestion points: The listMembers, auditSeats, and auditRepls functions in SKILL.md ingest JSON data from the Replit API.
- Boundary markers: No explicit boundary markers or instructions to ignore embedded content were found in the code snippets.
- Capability inventory: The skill has the capability to execute Bash commands (curl), write files (team-activity.csv), and perform network operations.
- Sanitization: No specific sanitization or validation of the API response data was observed in the provided snippets.
Audit Metadata