The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local Python scripts (script.py) to perform various utility tasks such as capability detection, workflow appending, and debrief scaffolding. These scripts are invoked via the command line using python3 and are intended for project lifecycle management.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface as it ingests data from external files such as .taskmaster/docs/prd.md and .taskmaster/tasks/tasks.json to inform its logic.
Ingestion points: .taskmaster/docs/prd.md, .taskmaster/tasks/tasks.json, and .taskmaster/reports/task-complexity-report.json are read into the agent context in Steps 1 and 2.
Boundary markers: None explicitly defined in the skill instructions to delimit the content of these files.
Capability inventory: The agent can execute shell commands via python3, invoke MCP tools (mcp__plugin_prd_go), and utilize built-in commands like /goal.
Sanitization: There is no mention of sanitization or validation of the content within the ingested PRD or task files before processing.
[DATA_EXFILTRATION]: The skill provides a link to https://atlas-ai.au/pro as an upgrade path for premium features. This is an informational URL related to the tool's commercial offering and does not involve the transfer of sensitive data.

handoff