The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructions direct users to clone a repository from an external GitHub account (github.com/alphaparkinc). This source is not recognized as a trusted organization or well-known service, representing a potential supply-chain risk.
[COMMAND_EXECUTION]: The documentation requires the execution of shell commands such as git clone and pip install. If the external source were compromised, these commands could facilitate the execution of malicious scripts during the installation process.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from local files (email templates) which can influence the agent's behavior if malicious instructions are embedded in the marketing copy.
Ingestion points: Content is read from local file paths provided to the test_newsletter_template function (SKILL.md).
Boundary markers: There are no delimited boundary markers or explicit instructions provided to the agent to ignore instructions found within the ingested file content.
Capability inventory: The skill has the capability to read local files (open) and perform network operations (requests via the MarketingEmailSpamPreventClient).
Sanitization: No sanitization or content filtering is implemented for the text read from external files before processing.
[DATA_EXFILTRATION]: The skill includes logic to read local files using open(template_path, 'r'). If an agent is directed to process a sensitive path (e.g., credentials or configuration files), the content could be read and potentially sent to the external analysis API (api.genpark.ai).

genpark-marketing-email-spamprevent-skill