The Agent Skills Directory

[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it interpolates raw article content into the Gemini analysis prompt without sanitization or protective delimiters. A malicious Markdown file could potentially include instructions designed to override the agent's behavior during the analysis phase.
Ingestion points: The scripts/process_goldmine.py script reads the body content of user-provided Markdown files.
Boundary markers: None detected; article content is appended directly to the prompt template string in the build_prompt function.
Capability inventory: The skill has the ability to execute the gemini CLI tool and write analysis results as Markdown files to the local filesystem (with paths restricted to the project root).
Sanitization: No escaping, validation, or filtering is performed on the article content before it is interpolated into the system prompt.
[EXTERNAL_DOWNLOADS]: The scripts/bootstrap.sh script automates the installation of the gemini-cli tool using official package managers like npm (@google/gemini-cli) and Homebrew. These represent well-known and trusted software distribution services.
[COMMAND_EXECUTION]: The skill utilizes shell scripts to manage the execution environment and automate article processing. This includes checking for and installing Python 3 and system-level dependencies via package managers, which may involve the use of sudo for administrative installation tasks.

content-goldmine-gemini