The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests and analyzes data from external URLs, creating a surface for potential indirect prompt injection attacks.
Ingestion points: External content is pulled from user-provided URLs and brand mentions via the scripts/dfs_client.py script, which is then processed by the agent to evaluate E-E-A-T signals.
Boundary markers: The process steps do not define explicit delimiters or boundary markers to distinguish between the skill's instructions and the untrusted content being analyzed, which could lead to the agent inadvertently following instructions found on the target pages.
Capability inventory: The skill executes local Python scripts (dfs_client.py, report_publish.py) and writes generated audit files to the local file system (SEO reports directory).
Sanitization: There is no mention of sanitization or filtering of the fetched web content before the agent evaluates it for Experience, Expertise, Authoritativeness, and Trust.
[COMMAND_EXECUTION]: The skill automates its workflow by executing local Python scripts through shell commands that incorporate user-supplied inputs such as domains and URLs. While this is the intended mechanism for the audit, it represents a tool-based capability that handles external data.

eeat-review