task-harness
Pass
Audited by Gen Agent Trust Hub on Sep 19, 2026
Risk Level: SAFECOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill workflow requires the agent to execute a shell script (
bash init.sh) to initialize the environment and restore context. It also instructs the agent to use Git commands (git status,git log,git commit,git push) for version control and progress tracking. - [DYNAMIC_EXECUTION]: The
init.shscript utilizes inline Python commands (python3 -c) to parse JSON data from thefeature_list.jsonfile to calculate and display task completion statistics. - [INDIRECT_PROMPT_INJECTION]: The skill ingests user-provided requirements and analyzes existing codebases to generate structured task lists. While these generated tasks are subsequently read by the agent, the risk is mitigated by the instruction to use rigid JSON templates, which limits the agent's tendency to interpret embedded instructions as primary directives.
- [DATA_EXPOSURE]: The initialization script accesses and displays project metadata, including local directory paths, Git commit history, and the status of local changes, to provide the agent with necessary context.
Audit Metadata