merge-agent-work
Pass
Audited by Gen Agent Trust Hub on May 5, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill includes a bash script that performs git operations by calling a function from an internal utility script located at
../utils/git-worktree-utils.sh. - [PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection because it accepts a user-provided parameter (the agent ID or timestamp) and passes it directly to a shell function.
- Ingestion points: Untrusted data enters the agent context via the first command-line argument assigned to
AGENT_IDinSKILL.md. - Boundary markers: No delimiters or instructions are present to prevent the agent from misinterpreting potential instructions in the input.
- Capability inventory: The skill has the capability to execute shell commands and modify the git state of the repository via the
merge_agent_workfunction inSKILL.md. - Sanitization: No escaping or validation is performed on the input in
SKILL.mdbefore it is passed to the shell function.
Audit Metadata