The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill defines a tool to read external email content, creating a vulnerability surface where malicious instructions in emails could influence agent behavior.\n
Ingestion points: Gmail message content via the gmail_read tool described in SKILL.md.\n
Boundary markers: The instructions do not define specific delimiters or instructions to the agent to treat email content as untrusted data.\n
Capability inventory: The agent has access to powerful tools such as run_code, write_file, and install_pkg as listed in the skill frontmatter.\n
Sanitization: The skill performs basic sanitization by stripping HTML and truncating body text to 500 characters.\n- [EXTERNAL_DOWNLOADS]: The skill installs the googleapis library via npm. This is a well-known official dependency for Google service integration and is considered a safe source.\n- [DYNAMIC_EXECUTION]: The skill generates tool logic in src/tools/gmail.ts at runtime based on the provided instructions. This mechanism of script generation and registration is a standard form of dynamic execution for adding capabilities.

add-gmail