board-of-directors
Pass
Audited by Gen Agent Trust Hub on Mar 29, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes
run_shell_commandinSKILL.mdto create project-specific directories (mkdir -p) for storing board session metadata and deliberation logs. - [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection (Category 8) as it processes external, untrusted 'proposals' across a multi-step deliberation chain.
- Ingestion points: The
proposalcontent enters the agent's context inSKILL.mdand is passed to five separate director personas. - Boundary markers: The skill lacks formal isolation (e.g., XML tags or specific delimiters) for the proposal content, relying on simple labels like
PROPOSAL: ${proposal}. - Capability inventory: The skill has the capability to execute shell commands (
mkdir) and write files (write_file) based on the outcome of the deliberation. - Sanitization: There is no evidence of input sanitization or validation performed on the proposal text before it is interpolated into the expert profiles' prompts.
Audit Metadata