learnings
Pass
Audited by Gen Agent Trust Hub on Oct 17, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATIONPRIVILEGE_ESCALATIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill features an attack surface by ingests untrusted data from historical logs and transcripts (located via runtime context or user-provided paths). It mitigates this risk through a clear boundary instruction that historical content must be treated as evidence rather than executable commands.
- [DATA_EXFILTRATION]: While the skill accesses project-specific logs which may contain sensitive information, it includes mandatory instructions to omit secrets and private content from the final chat report.
- [PRIVILEGE_ESCALATION]: The skill instructions specifically forbid the agent from changing file permissions or conducting indiscriminate filesystem scans to gain access to logs.
- [COMMAND_EXECUTION]: The skill enforces a strict read-only mode, explicitly prohibiting the re-execution of commands, scripts, or database queries discovered during the review of historical records.
Audit Metadata