skills-qa
Pass
Audited by Gen Agent Trust Hub on Jul 3, 2026
Risk Level: SAFE
Full Analysis
- [PROMPT_INJECTION]: The skill contains numerous strings associated with prompt injection, such as 'ignore previous instructions', 'disregard the above', and 'switch to developer mode'. These are explicitly used as detection patterns to audit other skills for malicious behavior and are not intended to override the agent's current instructions.
- [DATA_EXFILTRATION]: The instructions reference sensitive file paths, including
~/.ssh/,~/.aws/, and~/.config/gh/. These references are solely for a 'heuristic scan' meant to flag if other skills attempt to access these paths. The skill itself does not perform any file read or write operations on these sensitive directories. - [COMMAND_EXECUTION]: The skill mentions shell and code execution patterns like 'eval strings' and 'curl scripts'. These are documented as risk indicators to be flagged during the audit process of third-party skills, rather than commands to be executed by the skill itself.
- [PROMPT_INJECTION]: An indirect prompt injection surface is present because the skill is designed to ingest and analyze untrusted, attacker-controlled text (external SKILL.md files). The instructions acknowledge this risk and note that safety relies on the external execution environment (read-only subagent mode).
Audit Metadata