The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill instructs the agent to execute a remote shell script using the curl | bash pattern from https://parallel.ai/install.sh. This is a dangerous execution vector that allows for unverified code to run with the user's shell permissions.
[COMMAND_EXECUTION]: The skill makes extensive use of the parallel-cli tool to perform web operations. Several references (data-enrichment.md, deep-research.md, web-extract.md, web-search.md) interpolate user-provided $ARGUMENTS directly into shell command strings without visible sanitization, creating a significant risk for shell command injection.
[CREDENTIALS_UNSAFE]: The skill directs the agent to check the .env file in the project root for a PARALLEL_API_KEY and load it. While using environment files is common, instructing an AI to proactively search for and read credentials from sensitive local files increases the risk of accidental exposure or exfiltration.
[EXTERNAL_DOWNLOADS]: The skill installs software from external sources, including a shell script from parallel.ai and Python packages (python-dotenv, parallel-web-tools) from public registries.

parallel-web