The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes raw livestream transcript text as its primary data source. This untrusted content is passed to multiple sub-agents to generate recaps, Q&A entries, and lecture notes without explicit security boundaries or instructions to ignore embedded commands.
Ingestion points: Transcript files are ingested in SKILL.md and forwarded to agents in Step 4.
Boundary markers: No explicit delimiters or instructions to ignore embedded commands are present in the agent prompts when handling the transcript data.
Capability inventory: The skill possesses the ability to execute file system commands and interact with external APIs via Google Workspace and Kit MCP tools.
Sanitization: No specific sanitization or filtering is performed on transcript content before it is processed by the AI sub-agents.
[COMMAND_EXECUTION]: Shell commands are used for project file organization and archival tasks.
Evidence: SKILL.md and agents/training-doc-admin.md use cp, mv, and mkdir to manage transcript files based on titles and dates, though renaming logic includes slugification as a basic mitigation.
[DYNAMIC_EXECUTION]: The skill provides Python code snippets for the agent to execute at runtime for tasks not covered by existing tools.
Evidence: Step 9.6 of SKILL.md includes Python code for roster conversion (Excel to Google Sheets) and Google Drive shortcut creation, intended for execution by the agent.
[EXTERNAL_DOWNLOADS]: The skill integrates with external platforms through MCP for legitimate synchronization.
Evidence: Uses MCP tools for Google Workspace (Docs, Sheets, Drive) and Kit to sync processed content and subscriber data. These integrations target well-known services consistent with the skill's primary function.

training-process-video