The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses shell commands to invoke various Python scripts (both from local files and heredocs) to perform its core logic, including data retrieval and reporting.
[DYNAMIC_EXECUTION]: The skill generates and executes Python code at runtime using shell heredocs within the SKILL.md file to manage data flow. Furthermore, the provided scripts/fetch.py file uses ssl._create_unverified_context() to bypass SSL certificate validation for its network requests. While this is often done for compatibility with local environments that have outdated certificate stores, it disables standard protections against man-in-the-middle (MITM) attacks.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external content, such as package descriptions from the npm registry and user bios from GitHub profiles, to generate lead briefs and suggested outreach messages.
Ingestion points: External data is ingested from registry.npmjs.org and api.github.com within scripts/fetch.py and the embedded Python blocks in SKILL.md.
Boundary markers: The skill instructions include a 'Critical rule' mandating that output must strictly follow the API responses, and it includes a self-QA step to verify that generated content remains faithful to the sourced data.
Capability inventory: The skill has the capability to execute Python code, write files to the local system (e.g., /tmp/ and docs/), and make network requests to the npm and GitHub APIs.
Sanitization: Package names are URL-encoded before being interpolated into API request URLs to prevent path traversal or URI manipulation.

npm-downloads-to-leads