debian-linux-triage
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill incorporates untrusted user data from
${input:ProblemSummary}and${input:Constraints}into the prompt to generate remediation commands. This creates a potential surface where malicious instructions in the problem summary could influence the generated output.\n - Ingestion points: User-provided input variables in
SKILL.md.\n - Capability inventory: Generates system administration commands for
apt,systemctl,journalctl, anddpkg.\n - Boundary markers: No clear delimiters are used to separate user input from system instructions.\n
- Sanitization: The skill does not provide instructions to sanitize the input data.\n- [SAFE]: The functionality and tools requested are consistent with the skill's stated purpose of providing Linux system administration support.
Audit Metadata