The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-provided keyword lists and cluster maps to generate a publishing schedule. This ingested data is subsequently used to name files and as input for local scripts. The instructions do not specify sanitization or boundary markers for this external data, which could allow maliciously crafted keywords to influence agent behavior in the reporting phase.
[COMMAND_EXECUTION]: The skill relies on local Python scripts (scripts/dfs_client.py and scripts/report_publish.py) to fetch data and generate branded reports. The agent is instructed to run these scripts via the shell, passing keyword data and filenames as arguments. While this is the intended function, it represents the primary capability used to process external content.

content-calendar