The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill instructions direct the agent to prompt the user for their email and password and then execute a shell command including these credentials as plain-text arguments (npx visantlabs login -e EMAIL -p PASSWORD). This practice exposes sensitive credentials to shell history and system process lists. Similarly, configuring the MCP via claude mcp add places the API key directly in a command-line header.
[COMMAND_EXECUTION]: The setup workflow involves executing shell commands to read a sensitive local file (~/.visant/credentials.json) and to configure the environment, accessing private configuration paths that should be handled with caution.
[EXTERNAL_DOWNLOADS]: The skill requires the use of npx visantlabs, which fetches and executes code from the npm registry at runtime. While this is the service's own tool, it represents a dependency on external code execution.
[INDIRECT_PROMPT_INJECTION]: Tools such as document-extract and brand-guidelines-ingest allow the agent to process content from arbitrary URLs and PDF files. This data is incorporated into the agent's context, potentially allowing malicious instructions in the external data to influence the agent.
Ingestion points: document-extract (URL and PDF modes), brand-guidelines-ingest (external URLs or text).
Boundary markers: There are no instructions for using delimiters or warnings to mitigate the risk of instructions embedded in the external content.
Capability inventory: The skill performs network operations to api.visantlabs.com, reads local PDF files via pdf_path, and accesses the user's ~/.visant/credentials.json file.
Sanitization: No methods for sanitizing or validating the ingested external data are described before the content is returned to the agent context.

visant-mcp-connect