The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill's setup documentation instructs the agent to download and execute a remote shell script from https://parallel.ai/install.sh by piping the output of curl directly into bash.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted external data from the web, creating a vulnerability to indirect prompt injection.
Ingestion points: Processes web search excerpts (references/web-search.md), full-page content extractions (references/web-extract.md), and deep research reports (references/deep-research.md).
Boundary markers: There are no explicit delimiters or instructions provided to the agent to treat ingested web content as untrusted data.
Capability inventory: The skill possesses network access and file system write capabilities through the parallel-cli tool (e.g., via the -o and --target flags used to save results).
Sanitization: No sanitization, filtering, or validation is performed on the content retrieved from external URLs before it is provided to the agent.
[COMMAND_EXECUTION]: The skill templates interpolate user-supplied arguments ($ARGUMENTS) and other variables directly into shell command strings for the parallel-cli across all reference files.
[EXTERNAL_DOWNLOADS]: The skill initiates the download and installation of external packages, including parallel-web-tools and python-dotenv, from public registries.

parallel-web