The Agent Skills Directory

[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it processes content from external or user-provided sitemap files via the analysis script. Malicious instructions hidden in URL paths or metadata within a sitemap could be processed by the agent during the audit phase.
Ingestion points: The scripts/sitemap_analyzer.py script reads data from remote URLs or local file paths and prints the results for agent processing.
Boundary markers: Absent. There are no instructions or delimiters designed to protect the agent from interpreting embedded instructions in the sitemap content.
Capability inventory: The agent can execute a bundled Python script, perform network requests, and read the local filesystem.
Sanitization: Absent. The script performs structural parsing but does not validate or sanitize text content for potential malicious instructions.
[EXTERNAL_DOWNLOADS]: The scripts/sitemap_analyzer.py utility uses the standard urllib library to fetch sitemaps from remote URLs. This allows the skill to communicate with arbitrary external domains during the auditing process.
[COMMAND_EXECUTION]: The skill instructs the agent to run a bundled Python script that accepts arbitrary file paths or URLs as command-line arguments. This creates a surface for potential unauthorized file exposure if the agent is manipulated into auditing sensitive system files (e.g., credentials or configuration) instead of standard sitemaps.

site-architecture