doca-dpa-hl-tracer
Pass
Audited by Gen Agent Trust Hub on Oct 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to ingest and analyze DPA-side execution traces (binary trace files and human-readable logs). This creates a surface for indirect prompt injection, as an attacker controlling the DPA-side workload could potentially embed malicious instructions within the trace output to influence the agent's subsequent actions.
- Ingestion points:
bin_fileandlog_fileprocessed inTASKS.mdduring the capture, decode, and rendering pipeline. - Capability inventory: The skill involves executing the
doca_dpa_hl_tracerCLI and requires various system-level information (device names, ELF files, JSON configurations). - Boundary markers: The skill encourages quoting events verbatim but does not specify robust boundary markers or sanitization procedures for the potentially untrusted trace content.
- Sanitization: No explicit content validation or escaping is recommended before the agent processes the trace data.
Audit Metadata