steve-jobs-perspective
Pass
Audited by Gen Agent Trust Hub on Nov 15, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTIONTIME_DELAYED_CONDITIONAL
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands (e.g.,
git ls-remote) to check for updates from its official GitHub repository.- [EXTERNAL_DOWNLOADS]: The README recommends installation vianpx, which fetches the skill package from the author's public repository. It also suggests usinggit pullfor updates, leading to the execution of remote code.- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from user queries about product designs and business strategies. It lacks strict delimiters for this input, creating a potential surface for indirect prompt injection, although its capabilities are limited to maintenance and persona-driven analysis. - Ingestion points: User-provided product descriptions and business queries processed within
SKILL.md. - Boundary markers: Absent; relies on natural language instructions.
- Capability inventory: Shell command execution (
git), network access to GitHub, and local file writing for update tracking. - Sanitization: Absent.- [TIME_DELAYED_CONDITIONAL]: The update check operation is gated by conditional logic based on the date of the last check stored in a local state file (
.last-update-check), triggering only after specific intervals (30 days or 7 days).
Audit Metadata