The Agent Skills Directory

[PROMPT_INJECTION]: The skill instructions do not contain any patterns attempting to override system behavior, bypass safety filters, or extract system prompts. The language is purely instructional and focused on the technical usage of the dlt library.
[DATA_EXFILTRATION]: No patterns for exfiltrating sensitive data were found. The documentation correctly identifies best practices for secret management using secrets.toml and environment variables. Examples for connection strings use descriptive placeholders rather than hardcoded credentials.
[REMOTE_CODE_EXECUTION]: Code examples recommend standard installation via pip from PyPI. The provided maintenance script (scripts/check-updates.py) fetches package metadata from the official PyPI API using urllib, which is a standard and safe operation for version checking.
[OBFUSCATION]: The skill was analyzed for Base64 encoding, zero-width characters, homoglyphs, and other encoding tricks. No obfuscated content or hidden URLs were detected.
[INDIRECT_PROMPT_INJECTION]: The skill describes tools for ingesting data from REST APIs and databases. While this creates an attack surface where external data could influence an agent, the library emphasizes structured data handling, schema inference, and schema contracts, which mitigate risks associated with untrusted data.
[DYNAMIC_EXECUTION]: The library supports Python-based transformations (maps and filters). The documentation focuses on standard functional programming patterns without using unsafe dynamic execution primitives like eval() on external input.

dlt