The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill performs a global installation of the deepline package using a custom registry at https://code.deepline.com/api/v2/npm/. This involves downloading and executing software from an external third-party domain that is not a recognized trusted source.
[INDIRECT_PROMPT_INJECTION]: The skill retrieves lead data via the dropleads_search_people tool and incorporates fields like linkedin_url directly into a subsequent deepline enrich command. This pattern allows data from an external API to influence the execution of local tools.
Ingestion points: Data is collected from the output of the search tool and saved to deepline/data/quickstart_search.json before being processed into a CSV.
Boundary markers: No explicit delimiters or instructions are provided to the agent to treat the external tool output as potentially unsafe data.
Capability inventory: The skill has the capability to install software via npm, run shell commands, and execute Python scripts.
Sanitization: Only basic string cleaning (whitespace stripping and protocol normalization) is performed, which does not protect against instructions embedded in the data.
[DYNAMIC_EXECUTION]: A Bash HEREDOC is used to define and execute an inline Python script at runtime. This script is used to parse JSON data and generate CSV files, representing a runtime code generation and execution pattern.
[COMMAND_EXECUTION]: The skill invokes several high-impact shell commands, including global package installation, environment path modification, and authentication sequences that interact with a remote service.

deepline-quickstart