The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits a vulnerability to indirect prompt injection due to its use of untrusted user input in a persistent storage workflow.\n
Ingestion points: User inputs describing meeting translation needs are captured and processed.\n
Boundary markers: There are no explicit delimiters or instructions to treat data from the learnings.md file as untrusted when it is read back into the agent's context.\n
Capability inventory: The skill is instructed to read from and append new findings to references/learnings.md, creating a feedback loop where user-provided content can influence future agent behavior.\n
Sanitization: The skill lacks mechanisms for sanitizing or validating content before writing it to the local knowledge base.\n- [COMMAND_EXECUTION]: The skill documentation provides instructions for installing related tools through the command line.\n
Evidence: The 'Related skills' section lists several npx skills add commands for the sales-skills/sales repository to facilitate installation of complementary features.\n- [EXTERNAL_DOWNLOADS]: The skill references external resources for software installation and platform support.\n
Evidence: Links to official download paths and support contacts at transyncai.com are provided within the troubleshooting and setup sections.

sales-transync-ai