The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill incorporates an attack surface for indirect prompt injection by design, as it processes both user-supplied data and external reference files to generate strategic reports.
Ingestion points: The skill ingests user input regarding business type, goals, and budget, and reads external reference files located at ../ads/references/budget-allocation.md and ../ads/references/conversion-tracking.md to inform its planning logic.
Boundary markers: There are no explicit delimiters or protective instructions provided to the agent to distinguish between the structural instructions of the skill and potentially malicious instructions embedded within the user data or reference files.
Capability inventory: Across all scripts and instructions, the skill uses file-read capabilities and output generation to create various strategy documents (e.g., ADS-STRATEGY.md, BUDGET-PLAN.md). It does not possess high-risk capabilities like network exfiltration or shell execution.
Sanitization: The skill lacks explicit sanitization or validation steps for the content it reads, relying on the model's standard reasoning to process the data.
[SAFE]: The skill operates entirely within the scope of its stated purpose. All referenced assets are local to the package or within the anticipated workspace structure, and no suspicious network calls, obfuscation, or persistence mechanisms were found.

ads-plan