The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external sources which could potentially contain malicious instructions designed to influence the agent's behavior during automated analysis or keyword creation.
Ingestion points: The skill ingests search terms from the Apple Search Ads API (asa search-terms list) and competitor keyword data from Market Intelligence (asa competitors summary), as documented in references/playbooks/search-term-harvesting.md and references/playbooks/keyword-opportunity.md.
Boundary markers: Every write operation requires an explicit mutation proposal and a human-in-the-loop confirmation (requires_confirmation: true), as defined in the Analysis and Proposal Contract in SKILL.md.
Capability inventory: The skill possesses capabilities to modify account state, including adding keywords (asa keywords add), updating bids (asa keywords update), and managing negative keywords (asa negative-keywords add), primarily handled in references/asa-management.md.
Sanitization: External strings are normalized using Unicode normalization, trimmed of whitespace, and lowercased before comparison to prevent schema confusion and basic injection attempts, as specified in references/playbooks/keyword-opportunity.md.
[COMMAND_EXECUTION]: The skill executes shell commands via the adapty CLI to interact with Apple Search Ads services.
Evidence: The preflight script in SKILL.md executes adapty asa whoami to verify connectivity and credentials.
Context: This is the primary intended purpose of the skill. The script includes diagnostic logic to handle specific sandbox certificate errors by conditionally setting NODE_USE_SYSTEM_CA=0. It explicitly forbids the use of sudo and provides clear error routing for authentication and network issues.

apple-ads