b2c-production-triage
Pass
Audited by Gen Agent Trust Hub on Sep 25, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted external data such as production logs, job output, and business records via tools like
logs_get_recent,scapi_search, andwebdav_get. While this introduces an attack surface where malicious content embedded in logs could attempt to influence the agent's behavior, the skill mitigates this by instructing the agent to use minimal excerpts and to redact sensitive information before producing a handoff. - [DYNAMIC_EXECUTION]: The skill leverages platform-native 'Code mode' and SCAPI execution tools (
codemode.describe,scapi_execute) to investigate failed jobs and inspect system states. These features are used according to the vendor's intended operational workflows for incident triage rather than for unauthorized code injection. - [DATA_EXFILTRATION]: The skill provides clear security boundaries, explicitly instructing the agent to exclude tokens, payment details, and customer data from any handoffs or support cases. It directs the agent to keep unredacted evidence within the customer's approved storage rather than in public repositories.
Audit Metadata