The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the Bash tool to execute framework-specific binaries (e.g., gstack-skill-start, gstack-ios-qa-regen, gstack-skill-end) located in ~/.claude/skills/gstack/bin/. These executions are standard for the skill's operation within the gstack ecosystem.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes application source code, creating an attack surface for indirect prompt injection. However, it incorporates robust mitigations, including a 'User-origin gate' that ignores instructions in tool outputs or files unless they originate directly from the user's chat message.\n
Ingestion points: User application source code ($APP_SOURCE_DIR) processed in SKILL.md.\n
Boundary markers: The preamble instructions include a specific security boundary requiring that instruction blocks (GSTACK_INSTRUCTION_BEGIN) only be followed if they originate from the trusted gstack-skill-start tool result.\n
Capability inventory: The agent has access to Bash, Read, Write, and Edit tools to perform code regeneration and verification.\n
Sanitization: The generation process uses a specialized binary (gstack-ios-qa-regen) which performs structural validation on the input Swift code, rejecting non-standard types.

ios-sync