The Agent Skills Directory

Indirect Prompt Injection (HIGH): The skill is designed to 'Check out the current project state first (files, docs, recent commits)' to inform its brainstorming process. This is a significant vulnerability surface where malicious instructions hidden in the codebase or commit history could hijack the agent's behavior.
Ingestion points: Local project files, documentation, and git commit history (specified in SKILL.md).
Boundary markers: Absent. The skill does not define delimiters to separate project data from its own instructions.
Capability inventory: Writing design documents to the file system (docs/plans/) and executing git commit operations. It also invokes other skills like superpowers:writing-plans.
Sanitization: None. The skill processes external project content without validation or filtering.
Command Execution (LOW): The skill explicitly directs the agent to 'Commit the design document to git'. While a legitimate development workflow, it grants the agent the ability to modify the repository history based on instructions that could be influenced by external project data.
Data Exposure (LOW): By design, the skill accesses all project files and documentation. While no external exfiltration is explicitly scripted, the exposure of sensitive local project information to the LLM context is inherent to its function.

brainstorming