The Agent Skills Directory

[DYNAMIC_EXECUTION]: The script scripts/budget-optimizer.sh performs significant path traversal to move three directory levels above its own location to calculate a root path. It then executes an external shell script meta-kit.sh from this computed directory. Executing code from paths outside the skill's distribution folder is a security concern as the agent is relying on external, unverified system scripts.\n- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze campaign-level and adset-level performance data from advertising platforms. If these campaigns contain maliciously crafted text in their names or metadata, they could attempt to influence the agent's behavior during analysis.\n
Ingestion points: Campaign and adset insights processed in the invocation steps specified in SKILL.md.\n
Boundary markers: Absent; the instructions do not define delimiters or specific 'ignore embedded instructions' warnings for the data being analyzed.\n
Capability inventory: The skill possesses command execution capabilities (bash) and the ability to write logs to workspace/brand/learnings.md.\n
Sanitization: No validation, filtering, or sanitization steps are defined for the campaign metadata before it enters the agent's context.\n- [COMMAND_EXECUTION]: The skill uses the bash tool to run local wrappers that execute further system scripts. This provides the agent with access to the underlying shell environment, which requires oversight when coupled with external data inputs.

budget-optimizer