init-effort-with-herdr
Pass
Audited by Gen Agent Trust Hub on Sep 28, 2026
Risk Level: SAFECOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill constructs shell commands using user-supplied variables such as and . Commands like
treehouse get --lease --lease-holder <effort>andgit -C <path> switch -c <branch>are vulnerable to command injection if the inputs contain shell metacharacters like semicolons, backticks, or pipes. - [DYNAMIC_EXECUTION]: The skill generates bash script blocks from predefined templates for execution in the terminal. The lack of validation for the variables interpolated into these scripts represents a dynamic execution risk, as the generated script content is determined by unvalidated user input at runtime.
- [INDIRECT_PROMPT_INJECTION]: The skill takes a user-provided idea and interpolates it directly into a prompt sent to a secondary agent session via
/grill-with-docs <the idea>. This lacks boundary markers or instructions to prevent the input from potentially hijacking the downstream agent's behavior. - Ingestion points: The project idea is captured via user input in the SKILL.md execution flow.
- Boundary markers: None present; user input is directly concatenated into the agent prompt string.
- Capability inventory: The skill executes local CLI tools (git, treehouse, herdr) and spawns new agent sessions.
- Sanitization: No evidence of input validation or escaping for the user-provided text.
Audit Metadata