The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest untrusted data from external sources, which presents a risk of indirect prompt injection.
Ingestion points: The skill uses webfetch to retrieve live <title> and <meta name="description"> tags from user-provided URLs in SKILL.md.
Boundary markers: Absent. There are no instructions or delimiters provided to the agent to treat the fetched metadata as untrusted or to ignore embedded instructions.
Capability inventory: The skill possesses the capability to execute shell commands (python scripts/dfs_client.py and python scripts/report_publish.py), write files to the local filesystem (METADATA-<domain>-<date>.md), and generate PDF/HTML reports.
Sanitization: Absent. The instructions do not specify any sanitization, escaping, or validation of the fetched web content before it is processed or included in generated reports.
[COMMAND_EXECUTION]: The skill triggers the execution of local Python scripts as part of its core workflow.
Evidence: The skill invokes python scripts/dfs_client.py and python scripts/report_publish.py using arguments derived from user input (domain, keyword, report filename).
Risk: If the arguments passed to these scripts (like the domain or keyword) are derived from the untrusted web content fetched in the audit phase without sanitization, it could lead to command injection vulnerabilities.

metadata-optimizer