The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local Python scripts (scripts/dfs_client.py and scripts/report_publish.py) to perform audits and generate client reports. These scripts are invoked via the shell, with user-supplied URL inputs passed as arguments.
[INDIRECT_PROMPT_INJECTION]: The skill fetches data from external, untrusted URLs using webfetch and a parsing script to analyze canonical tags. This data is then used to generate markdown and HTML reports. An attacker-controlled website could include malicious instructions designed to manipulate the agent's analysis or the resulting report content.
Ingestion points: External web content retrieved from user-specified URLs in webfetch and dfs_client.py calls.
Boundary markers: None present; the instructions do not specify delimiters or warnings to ignore embedded instructions in the fetched HTML.
Capability inventory: File writing (CANONICAL-<domain>.md) and local script execution (python scripts/...).
Sanitization: No explicit sanitization or validation steps are defined for the data retrieved from external sources before it is processed or used in report generation.

canonical-review