The Agent Skills Directory

[COMMAND_EXECUTION]: The skill is configured to execute a specific shell script (setup-superpower-loop.sh) located within the plugin's root directory. It passes parameters, including paths resolved from user arguments, to this script to initialize a continuous execution loop.
[PROMPT_INJECTION]: The skill contains a vulnerability surface for indirect prompt injection because it reads and processes external data from local Markdown files and incorporates that content into prompts for subagents without sanitization.
Ingestion points: The skill reads _index.md and various task-*.md files from the docs/plans/ directory to define the scope and logic of tasks.
Boundary markers: While the skill uses Markdown headers (e.g., ## Task Assignment) to separate content, it lacks explicit defensive delimiters or instructions to the LLM to disregard potentially malicious instructions embedded within the task files.
Capability inventory: The skill has the ability to manage tasks via TaskUpdate, spawn subagents using the Agent tool, and execute restricted shell commands via Bash.
Sanitization: There is no evidence of validation or sanitization of the content extracted from the plan files before it is interpolated into the mandatory prompt templates used for subagent assignments.

executing-plans