The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted data from the application under test (such as UI text, network logs, and system alerts) which serves as an ingestion surface for potential indirect prompt injection.
Ingestion points: SKILL.md and references/bug-triage.md (via snapshot -i, network dump, and logs path commands).
Boundary markers: The skill does not specify explicit delimiters for data retrieved from the testing environment.
Capability inventory: The skill utilizes the agent-device CLI via shell execution (Bash).
Sanitization: The instructions proactively mitigate risks by directing the agent to use tool-generated reference identifiers (e.g., @e12) instead of raw text labels, explicitly noting this avoids shell-quoting hazards.
[EXTERNAL_DOWNLOADS]: The documentation references installation commands for the CLI and associated plugins from the author's official GitHub repository.
Evidence: README.md mentions npx -y skills add -y -g yigitkonur/skills-by-yigitkonur/skills/run-agent-device.
Context: These are vendor-owned resources used for skill deployment and do not involve unauthorized or suspicious third-party downloads.
[SAFE]: No evidence of credential theft, data exfiltration, obfuscation, or persistence mechanisms was found. The skill's behavior is consistent with its stated purpose of providing an iOS testing workflow.

run-agent-device