The Agent Skills Directory

[COMMAND_EXECUTION]: The analysis procedure includes a bash command that runs a Python script using a string template for the brand name. The placeholder [Brand_Name] is placed directly inside single quotes: brand = '[Brand_Name]'. If the agent replaces this with user-provided input containing single quotes (e.g., '; import os; os.system('id'); #), it would result in arbitrary code execution within the Python environment. This pattern of dynamic code generation from unvalidated input is a significant risk.
[PROMPT_INJECTION]: The skill is designed to fetch and process untrusted content from external sites (Reddit, YouTube, Wikipedia) using WebFetch. This creates a surface for indirect prompt injection, where malicious instructions embedded in brand mentions on those platforms could influence the agent's behavior during the analysis phase.
Ingestion points: WebFetch calls to YouTube, Reddit, and Wikipedia search results in Step 2 of the analysis procedure.
Boundary markers: Absent. The instructions do not define delimiters or provide 'ignore embedded instructions' warnings for the fetched content.
Capability inventory: Bash, Write, Read, WebFetch, Glob, Grep.
Sanitization: None specified; the agent is instructed to directly analyze the fetched content for sentiment and scoring without filtering.

geo-brand-mentions