The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from Twitter/X (keywords, author posts) which is exported to CSV files for further analysis. This represents a vulnerability surface where malicious instructions embedded in tweets could influence agent behavior during analysis.
Ingestion points: SKILL.md (via data fetched from the Xpoz API and saved to tweets.csv).
Boundary markers: Absent; there are no specific instructions or delimiters provided to the agent to distinguish between its instructions and the untrusted CSV content.
Capability inventory: The skill utilizes subprocess.run (shell execution), urllib.request (network operations), and package managers (pip, npm).
Sanitization: Absent; the skill suggests loading the data directly into pandas for analysis without prior filtering or escaping of content.
[COMMAND_EXECUTION]: The skill uses subprocess.run to dynamically configure the mcporter tool. It incorporates an access token retrieved via a network request to the vendor's OAuth endpoint (https://mcp.xpoz.ai/oauth/token). While the source is a vendor-owned resource, the pattern of constructing shell commands from network-retrieved data is documented here as a functional characteristic.
[EXTERNAL_DOWNLOADS]: The skill downloads and installs the xpoz library from the Python Package Index and the @xpoz/xpoz library from the npm registry. It also retrieves configuration and dataset files from the vendor's official infrastructure (xpoz.ai).

twitter-data-export