The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests content from external websites, which could contain malicious instructions designed to manipulate the agent's behavior or override its safety guidelines.
Ingestion points: Untrusted data is retrieved from the api.parallel.ai service via the search, research, and extract methods in scripts/parallel_web.py.
Boundary markers: The skill does not implement explicit boundary markers (such as XML tags or clear delimiters) or specific safety instructions to help the agent distinguish between search results and legitimate system instructions.
Capability inventory: The skill possesses the ability to write files to the local project directory (via the -o flag and Write tool) and execute shell commands (via the Bash tool).
Sanitization: There is no evidence of content filtering, escaping, or sanitization of the retrieved web data before it is presented to the agent.

alterlab-parallel-web