The Agent Skills Directory

[DATA_EXFILTRATION]: The skill documents the use of the --deliver webhook:<url> flag, which enables the agent to automatically POST command results to an external server. This creates a direct mechanism for exfiltrating sensitive CRM data, contact lists, or API keys retrieved via other commands.
[PRIVILEGE_ESCALATION]: The skill provides access to a suite of admin commands, including admin impersonate, which allows the agent to assume the identity of other users within the CRM system, effectively granting it super-admin permissions.
[EXTERNAL_DOWNLOADS]: The instructions facilitate the download and installation of the CLI binary from the author's infrastructure via NPM (@mvanhorn/printing-press-library) and GitHub (github.com/mvanhorn/printing-press-library).
[REMOTE_CODE_EXECUTION]: The skill triggers the execution of remote code during the prerequisite setup phase through npx package installation and go install from external source repositories.
[COMMAND_EXECUTION]: The skill is configured to execute a local binary and utilizes a which feature that interprets natural language into shell commands executed through the agent's Bash capabilities.
[PERSISTENCE]: The skill implements persistence by storing authentication tokens (auth set-token), configuration profiles (profile save), and local execution logs (feedback) in the user's home directory.
[INDIRECT_PROMPT_INJECTION]:
Ingestion points: The agent processes untrusted data from CRM sources, including chat messages (chat list-messages), email sequences, and public form submissions.
Boundary markers: There are no delimiters or explicit instructions provided to prevent the agent from following commands embedded within the CRM data.
Capability inventory: The agent has access to shell execution, administrative CRM actions, and network-based output delivery.
Sanitization: The skill provides no documentation of input validation or sanitization for content retrieved from external API endpoints.

pp-conduyt-crm