new-mission
Pass
Audited by Gen Agent Trust Hub on Sep 2, 2026
Risk Level: SAFECOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute a bash script to retrieve and format the current system time and IANA timezone for logging purposes. This script uses standard utilities like
date,readlink, andprintfto ensure accuracy. - [DYNAMIC_EXECUTION]: The skill provides instructions for optional integration with an external review tool (
ai-review.sh). It constructs a command path using the skill's directory structure to perform automated planning reviews. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data, including user-provided mission descriptions and records from third-party task/todo systems.
- Ingestion points: User input triggers and external task records (GitHub/Todo systems) as defined in Step 0 and Step 1.
- Boundary markers: The skill implements a robust 'human-in-the-loop' boundary, requiring a multi-step plan review and explicit user confirmation ('Go') before any actions are taken.
- Capability inventory: The agent can perform file system reads (scanning for task context), file writes (saving mission reports/prompts), and command execution (shell scripts/external reviews).
- Sanitization: The skill mandates a 'Self-Review' phase (Step 3) and a 'User Review' phase (Step 4) where the plan's alignment with user intent is verified before execution.
Audit Metadata