lifeguard-in-lanes
Pass
Audited by Gen Agent Trust Hub on Oct 1, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted data from external sources, specifically GitHub issue bodies and comments, which could be controlled by third parties.
- Ingestion points: Reads issue body, comments (including specific tags like
CHECK-IN,HANDOFF,STUCK,DONE), and PR details from GitHub using theghCLI as defined inSKILL.md. - Boundary markers: The instructions do not define clear delimiters or use "ignore instructions" wrappers when parsing the content of these comments, increasing the risk of the agent following instructions embedded within them.
- Capability inventory: The agent has the ability to send messages to other agent sessions (
SendMessage), push notifications to the user (PushNotification), and interact with the GitHub API (gh api,gh pr view). A successful injection could misuse these tools to spam users or exfiltrate state information. - Sanitization: No specific sanitization or filtering logic is described for the external content before it is used to determine the agent's next action.
- [COMMAND_EXECUTION]: The skill relies on the execution of shell-based commands (
ghCLI) to perform its monitoring tasks. While these tools are standard, the inputs (such as PR identifiers and commit SHAs) are retrieved from untrusted issue comments, which presents a surface for potential command injection if the underlying tool execution is not properly managed by the platform.
Audit Metadata