The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/probe_providers.py uses subprocess.run to execute /usr/bin/osascript. The commands executed are hardcoded in a mapping (QUERIES) and do not include any user-provided input, preventing command injection vulnerabilities. The AST flag for shell execution is a false positive as the script uses a list of arguments rather than a shell string.
[DATA_EXPOSURE]: The skill uses os.open with os.O_EXCL and 0o600 permissions when writing report files. This ensures that existing files are not overwritten (preventing symlink attacks or data destruction) and that the resulting output is only readable by the user who ran the script.
[PRIVILEGE_ESCALATION]: The instructions in SKILL.md explicitly forbid recommending high-risk actions such as disabling System Integrity Protection (SIP), resetting Transparency, Consent, and Control (TCC) settings, or using sudo to bypass system protections. It correctly identifies that these actions are often ineffective and dangerous.
[SAFE]: External links point to official Apple support documentation (support.apple.com). No unauthorized network activity, external data exfiltration, or obfuscated content was detected across the skill files.

apple-effects-reliability