The Agent Skills Directory

[SAFE]: The skill enforces a 'Human-in-the-loop' confirmation model for all mutations, requiring the agent to present a proposal to the user and obtain explicit permission before executing any command that spends money or modifies account state.\n- [SAFE]: A robust idempotency system is implemented using --idempotency-key. The skill provides specific instructions on minting unique keys per operation to ensure that network retries or ambiguous results do not lead to duplicate campaign creation or keyword additions.\n- [SAFE]: The preflight verification script is designed for environment diagnostic purposes (checking CLI version and authentication status). It uses safe shell practices, handles environment-specific certificate errors without requesting elevated privileges, and does not access sensitive system files or credentials.\n- [SAFE]: The skill instructions explicitly prohibit dangerous behaviors such as using 'sudo', resetting Keychains, or hand-rolling installation scripts, directing the user toward secure official setup flows instead.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from Apple Search Ads and competitor Market Intelligence. This risk is mitigated through instructions for mandatory Unicode normalization, whitespace trimming, and lowercase conversion of external strings, combined with the requirement for human verification of all findings before action is taken.

apple-ads