The Agent Skills Directory

[DYNAMIC_CONTEXT_INJECTION]: The skill executes shell commands (cat) at load time via the ! syntax to read local files such as .agents/product-marketing-context.md and dataslayer-config.json. These operations are limited to reading specific files within the skill's directory and do not involve sensitive system paths.
[DYNAMIC_EXECUTION]: The skill is designed to write and execute Python scripts at runtime to generate PDF reports. While this is the intended primary function, it involves dynamic code generation and execution based on a reference implementation and external data.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from marketing platforms (via MCP tools) and manual user inputs to populate reports. This ingestion surface could be exploited if the source data contains malicious content intended to influence the agent's behavior during report generation.
Ingestion points: Reads dataslayer-config.json, MCP tool outputs, and manual data (CSV/JSON) provided by the user.
Boundary markers: None explicitly defined to separate untrusted data from instructions.
Capability inventory: Includes capabilities for shell execution (Bash), package installation (pip), file writing (cat >), and network operations (via the requests library in generated code).
Sanitization: The skill includes logic to handle missing files but does not provide explicit instructions for sanitizing or escaping ingested text data.
[EXTERNAL_DOWNLOADS]: The skill installs well-known Python packages (reportlab, pillow, requests) from official registries and is capable of downloading image assets from user-provided URLs to include as logos in reports.

ds-report-pdf