agenttrace-session-audit
Pass
Audited by Gen Agent Trust Hub on Aug 28, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes AI session logs (e.g., JSON exports) which may contain untrusted content from previous agent interactions or external tool outputs. While the skill focuses on operational metrics, the underlying data source is external and could theoretically contain instructions designed to influence the agent's summary or audit report. * Ingestion points: Ingests session data from files or directories via the agenttrace command in SKILL.md. * Boundary markers: No explicit delimiters or sanitization instructions are provided for the input logs. * Capability inventory: Executes the agenttrace binary and cargo run via subprocess. * Sanitization: No explicit sanitization of log content is mentioned.
- [COMMAND_EXECUTION]: The skill executes the agenttrace CLI tool and includes a fallback to cargo run -q -p agenttrace -- if the binary is missing and the user is within the agenttrace source repository. These commands are used to analyze local session data and are consistent with the skill's observability purpose.
Audit Metadata