The Agent Skills Directory

[PROMPT_INJECTION]: The skill ingests external data from the Apple Search Ads API (e.g., campaign names, keywords, search terms), which creates a surface for indirect prompt injection if account content is manipulated.
Ingestion points: references/playbooks/account-health.md and references/playbooks/structure-audit.md through the use of adapty asa read commands.
Boundary markers: Present in the form of a 'Finding contract' and evidence mapping, though there are no explicit instructions to disregard commands embedded in the processed data.
Capability inventory: The skill is limited to read-only operations using the adapty CLI, with explicit prohibitions against creating, updating, or deleting account entities.
Sanitization: Basic text normalization is used for keyword comparison, but it does not specifically filter for malicious prompt instructions.
[SAFE]: The skill demonstrates a clear security design by enforcing a read-only boundary and directing users to specific write-capable skills for any required changes, minimizing the risk of accidental or unauthorized mutations.

apple-ads-audit