The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides complex shell command templates using curl, jq, grep, and bash logic (e.g., loops for Nitter instances) to be executed by the agent. These commands use variable interpolation for parameters like handles and post IDs, which could lead to command injection if the agent does not properly sanitize these inputs from the source URLs.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and normalize untrusted content from social media platforms, creating an attack surface for instructions embedded in social posts to influence the agent.
Ingestion points: Social media post text, metadata, and engagement data fetched from external URLs via shell commands or browser interaction (references/strategies.md).
Boundary markers: The instructions provide a JSON normalization schema but do not define delimiters or specific instructions for the agent to ignore embedded commands within the fetched content.
Capability inventory: The skill utilizes shell command execution, filesystem writes to ~/Documents/social-fetches/, and automated browser tools.
Sanitization: No sanitization, escaping, or validation of the fetched social media content is specified before it is returned to the agent context.
[EXTERNAL_DOWNLOADS]: The skill retrieves data from multiple third-party services and mirrors to bypass anti-bot protections, including api.fxtwitter.com, nitter.net, scrapecreators.com, and apify.com. These services receive potentially sensitive data such as post URLs and, in the case of paid scrapers, API keys retrieved from the environment.

social-fetch