The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The installation instructions include the command curl -s "https://code.deepline.com/api/v2/cli/install" | bash. This pattern is highly insecure as it executes arbitrary remote code with local shell privileges without verification.
[COMMAND_EXECUTION]: The skill executes local commands using the deepline CLI and runs an embedded Python script via python3 -c for data validation, including file reading and processing logic.
[EXTERNAL_DOWNLOADS]: The skill fetches content from and interacts with APIs hosted at code.deepline.com.
[DATA_EXFILTRATION]: The deepline auth register command indicates that the skill facilitates the handling and external transmission of authentication credentials to the vendor's servers.
[PROMPT_INJECTION]: The skill is vulnerable to Indirect Prompt Injection (Category 8) because it uses template interpolation (e.g., {{First Name}}, {{Company}}) to insert raw CSV data into CLI command arguments. This allows malicious strings in the input data to manipulate the generated command.
- Ingestion points: External CSV files like leads.csv.
- Boundary markers: None; data is directly inserted into JSON-like strings in the command line.
- Capability inventory: Execution of the deepline CLI, file modification via the --in-place flag, and network access for data enrichment.
- Sanitization: None detected; the skill relies on the raw values provided in the CSV columns.

waterfall-enrichment