The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted external data, such as user-provided campaign parameters and website content retrieved during account audits via the web_fetch tool. This creates a surface where malicious instructions could be embedded in the ingested data to influence the agent's actions.
Ingestion points: User campaign requests (e.g., in README.md) and external web content fetched for tracking verification and audits (referenced in references/audit.md).
Boundary markers: The skill does not define specific delimiters or instructions to ignore embedded commands within ingested data to isolate it from its core logic.
Capability inventory: The skill's configuration (in SKILL.md and agents/openai.yaml) includes access to powerful tools such as bash, run_command, web_fetch, write_to_file, and replace_file_content.
Sanitization: No explicit procedures for sanitizing or validating external content before it is processed by the agent are documented in the skill's playbooks.

paidads