skills/garrytan/gstack/ios-sync/Gen Agent Trust Hub

ios-sync

Pass

Audited by Gen Agent Trust Hub on Sep 17, 2026

Risk Level: SAFECOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes several local vendor utilities including gstack-skill-start, gstack-ios-qa-regen, and gstack-skill-end. These binaries, located in the .claude/skills/gstack/bin/ directory, are used to manage the skill lifecycle, perform specialized Swift code generation, and handle telemetry.
  • [DYNAMIC_EXECUTION]: Within the context recovery section, the skill utilizes eval to source the output of a local utility named gstack-slug. This is an operational requirement for the skill to establish project-specific environment variables for the agent.
  • [INDIRECT_PROMPT_INJECTION]: The skill's preamble instructs the agent to follow instruction blocks (GSTACK_INSTRUCTION_BEGIN) that may be returned by its initialization tool. This creates a potential surface for indirect instruction injection. However, the skill explicitly mandates a validation process where instructions are only followed if they strictly originate from the tool's immediate output and contain a matching SESSION_ID, mitigating risks from external or untrusted data sources.
  • [DATA_EXFILTRATION]: The skill reports session telemetry and operational outcomes to a vendor-managed infrastructure using the gstack-skill-end tool. This behavior is documented and tied to the legitimate functioning of the vendor-provided skill ecosystem.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 17, 2026, 01:04 PM
Security Audit — agent-trust-hub — ios-sync