The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes shell commands for API interaction and workflow logic, utilizing tools like ./secretcurl, jq, grep, and date.
[EXTERNAL_DOWNLOADS]: Communicates with the external API at api.admanage.ai to check spend data, create campaigns, and launch advertisements.
[DYNAMIC_EXECUTION]: Uses python3 -c to execute a small snippet of Python code for numeric comparisons within the daily spend circuit-breaker logic.
[INDIRECT_PROMPT_INJECTION]: The skill's behavior is driven by declarative configuration files (config.yaml and config.create.yaml) stored in the repository, which creates an attack surface for instructions embedded in the data.
Ingestion points: Reads skills/schedule-ads/config.yaml and skills/schedule-ads/config.create.yaml to determine ad targeting, creative content, and budgets.
Boundary markers: None identified in the prompt templates.
Capability inventory: Perform network requests via ./secretcurl, write state to .admanage-state/, and log activity to memory/logs/.
Sanitization: Validates the structural shape of the YAML files, ensures media URLs use the https:// protocol, and uses regex to verify that spend and cap values are numeric.

schedule-ads