The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill fetches and analyzes content from arbitrary external URLs (Phase 1: Discovery), which serves as an entry point for untrusted data. An attacker could embed malicious instructions on a website to influence the agent's behavior during an audit.
Ingestion points: Website HTML and sitemaps fetched via WebFetch or curl during the discovery phase.
Boundary markers: No explicit delimiters or instructions to ignore embedded commands are defined in the orchestration logic.
Capability inventory: The skill has access to powerful tools including Bash, Write, and WebFetch across its various components.
Sanitization: There is no evidence of sanitization or filtering of the fetched content before it is passed to subagents for analysis.
[COMMAND_EXECUTION]: The skill uses the Bash tool and executes a local Python script (generate_pdf_report.py) to generate PDF reports. While this is consistent with the skill's stated purpose, it represents a path for command execution that relies on the integrity of the local script environment.

geo