task-track
Pass
Audited by Gen Agent Trust Hub on Nov 5, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill automatically handles the installation of its Node.js dependencies (e.g., 'yaml') from the official npm registry using 'npm ci' with security flags like '--ignore-scripts' and '--no-audit' if missing at runtime.\n- [COMMAND_EXECUTION]: The skill frequently executes local system commands to perform its core functions. This includes using 'git' to read repository state and commit history, as well as launching external programs like browsers (Google Chrome) or terminal emulators to provide a user interface for the task workspace.\n- [INDIRECT_PROMPT_INJECTION]: The skill reads external Markdown files (tasks, initiatives, and areas) from the project directory. Because this content is then incorporated into the agent's context to inform task management decisions, there is a risk that malicious instructions embedded in a project's documentation or task descriptions could influence the agent's behavior.\n
- Ingestion points: Files are read from the configured 'teamArtifacts' path in the 'readProgress' and 'readItem' functions within 'scripts/task-track.cjs'.\n
- Boundary markers: Instructions in 'SKILL.md' emphasize treating record text as data, but the implementation does not utilize strict technical delimiters or sanitization of Markdown content before it is processed by the agent context.\n
- Capability inventory: The skill can execute shell commands, run a local web server, and write files to the 'tmp/' directory.\n
- Sanitization: The skill applies HTML escaping for the visualization UI, but raw record content is ingested directly for agent processing.\n- [DYNAMIC_EXECUTION]: The skill dynamically generates shell scripts (such as '.sh', '.cmd', or '.command' files depending on the OS) in the project's 'tmp/task-tracking/launch/' directory. These scripts are then executed to launch the local workspace server in a visible terminal window on the host machine.
Audit Metadata