price-change
Pass
Audited by Gen Agent Trust Hub on Oct 7, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONCREDENTIALS_UNSAFE
Full Analysis
- [COMMAND_EXECUTION]: The skill uses shell commands including git, grep, and cd to verify changes and push code directly to the main branches of the agenthub, psquared-websites, and inboxmate-docs repositories.
- [INDIRECT_PROMPT_INJECTION]: The pipeline ingests text from multiple external sources such as website HTML, documentation files, and comparison pages to plan and calculate pricing updates.
- Ingestion points: Website source files (index.html, i18n.js, agb.html) and documentation (usage-and-billing.mdx).
- Boundary markers: No explicit delimiters or instructions are provided to the agent to ignore potentially malicious embedded commands in the processed source files.
- Capability inventory: Accesses Stripe production via MCP (create_price, create_coupon) and performs git push operations to production repositories.
- Sanitization: The skill lacks validation or sanitization for the ingested data before it is used to plan production changes.
- [CREDENTIALS_UNSAFE]: The skill accesses local configuration files and environment variable templates (subscriptionUtils.ts, .env.example) that expose internal infrastructure details like Stripe price IDs and pricing structures. It also uses absolute paths tied to a specific local user profile (/Users/martinpammesberger/), which limits portability and reveals local environment structure.
Audit Metadata