The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests and processes real-time data from X (Twitter) through the xAI API in scripts/analyze.sh, scripts/digest.sh, scripts/search.sh, and scripts/trending.sh. Ingestion points: API responses from xAI containing synthesized Twitter search results. Boundary markers: The agent uses system prompts that enforce a strict JSON schema and prohibit markdown code fences to delimit output. Capability inventory: The skill possesses network access via curl to the xAI API and file system write access to cache/runs/. Sanitization: Data is handled using standard json parsing in Python, and environment variables are sanitized using shlex.quote in scripts/common.sh.
[COMMAND_EXECUTION]: The load_config function in scripts/common.sh parses a .env file and dynamically generates a temporary shell script for sourcing. While this involves dynamic command generation, the skill employs shlex.quote within a Python helper to escape values, preventing command injection from the configuration file.
[COMMAND_EXECUTION]: The functions resolve_accounts and resolve_topics in scripts/common.sh use the eval command to perform dynamic variable lookups based on a category identifier. This behavior is mitigated by the validate_category function, which strictly restricts category names to alphanumeric characters and underscores.

x-research