visa-acceptance-best-practices

Pass

Audited by Gen Agent Trust Hub on Sep 20, 2026

Risk Level: SAFE
Full Analysis
  • [CREDENTIALS_UNSAFE]: The skill follows security best practices for credential management. In references/activities/act_02_obtain-merchant-credentials.md, it explicitly instructs the user to store credentials in local.properties rather than hardcoding them. Furthermore, it proactively recommends creating a .claude/settings.json file to deny the AI agent permission to read sensitive files like local.properties and .env files.
  • [EXTERNAL_DOWNLOADS]: The skill references several external resources, all of which are from trusted or well-known sources. These include documentation from developer.visaacceptance.com, SDK dependencies from repo.visa.com, sample code from github.com/CyberSource, and the Gradle wrapper from services.gradle.org. These downloads are documented as functional requirements for setting up the payment integration.
  • [PRIVILEGE_ESCALATION]: In references/activities/act_00c_connectivity-precheck.md, the skill describes a remediation step for corporate SSL interception that involves using sudo keytool to import certificates into the JDK truststore. This is a legitimate administrative action for developers in restricted network environments and is presented as a manual fix for the user rather than an automated process.
  • [COMMAND_EXECUTION]: The skill uses shell commands via grep and sed for project analysis (e.g., detecting versions in build.gradle) and adb for interacting with PAX terminals. These commands are constrained to the skill's primary purpose of Android development and terminal integration.
  • [INDIRECT_PROMPT_INJECTION]: The skill possesses a potential attack surface as it ingests and modifies Android project files.
  • Ingestion points: Project configuration files such as settings.gradle, build.gradle, and AndroidManifest.xml are read during the planning phase (act_00_planning.md).
  • Boundary markers: Explicit boundary markers are not used, but the ingestion is limited to specific grep patterns for version and configuration detection.
  • Capability inventory: The skill has the capability to write to build files, update local.properties, execute ./gradlew build commands, and perform adb operations on connected devices.
  • Sanitization: The skill mitigates risks by requiring explicit developer approval for every plan change and version upgrade detected during analysis.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 20, 2026, 03:00 AM
Security Audit — agent-trust-hub — visa-acceptance-best-practices