The Agent Skills Directory

[SAFE]: The skill performs analytical tasks using predefined markdown templates and cost benchmarks. It does not invoke shell commands, access sensitive system files (like credentials or SSH keys), or perform network operations outside of standard connector lookups.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a data ingestion surface as it processes campaign parameters and connector data, though it lacks dangerous capabilities to act upon malicious instructions embedded in that data. 1. Ingestion points: Data enters the agent context via user-provided budget parameters and optional connectors such as social platform analytics, CRM, and influencer databases. 2. Boundary markers: The skill instructions require labeling projected metrics as Measured, User-provided, or Estimated, but do not specify technical delimiters to isolate untrusted data from the prompt. 3. Capability inventory: The skill is restricted to writing markdown planning documents to the memory/ directory; it has no access to subprocesses, dynamic code execution, or network tools. 4. Sanitization: No explicit sanitization or filtering logic is defined for the input data before it is interpolated into the output templates.

budget-optimizer