devops-agent-cost-insights
Pass
Audited by Gen Agent Trust Hub on Oct 12, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- Data Ingestion from External Logs: The skill ingests untrusted data from the AWS DevOps Agent journal and AWS CloudTrail events.
- Ingestion Points: Data enters the context via tools like
get_investigation_journal_records,get_tool_calls, andlookup_cloudtrail_eventsas described in Step 3 of SKILL.md. - Boundary Markers: The skill implements strict boundary markers by requiring records to match both the reporting window and the active role scope (Step 3: Attribution test).
- Capability Inventory: The skill uses
use_awsfor read-only operations but explicitly forbids paid operations likeStartQueryorGetMetricData(Zero-cost rule). - Sanitization: The instructions require the agent to parse serialized JSON in
CloudTrailEventand verify specific identity fields likesessionIssuer.arnbefore processing, which acts as a sanitization layer against potentially malicious log content.
Audit Metadata