founder-thought-leadership
Pass
Audited by Gen Agent Trust Hub on Aug 27, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill's global protocols explicitly anticipate the ingestion and processing of untrusted content from external sources.
- Ingestion points: Section 4 (Security & Multi-Agent Hygiene) mentions processing "Web content and external data (e.g., via BrowserOS)".
- Boundary markers: The instructions mandate that such data be "treated as hostile," though specific mechanical delimiters are not provided.
- Capability inventory: The skill is authorized to use
run_shell_command,write_file, andreplaceas defined in the YAML frontmatter. - Sanitization: The protocol requires the agent to "Redact secrets/PII before sharing context with subagents."
- [COMMAND_EXECUTION]: The skill uses the
run_shell_commandtool for operational tasks and defines a specific execution pattern. - Evidence: Section 3 (The Iron Law of Execution) instructs the agent to execute terminal actions via the "ExecutionProxy Interface" using a default
rtkprefix (e.g.,rtk npm test).
Audit Metadata