The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses shell commands (gh issue create, linear issue create, open, cat) with variables derived from user-provided feature descriptions and titles. While filenames undergo kebab-casing, the title string is directly interpolated into CLI arguments (--title "<title>"). If a user provides a title containing shell metacharacters or command substitution patterns (e.g., $(...)), it could lead to unintended command execution depending on the agent's shell environment.
[PROMPT_INJECTION]: The skill is vulnerable to Indirect Prompt Injection. It ingests untrusted data from the feature_description argument and from the content of existing brainstorm files (docs/brainstorms/*.md). This data is then used to guide the behavior of multiple research agents and the plan generation process.
Ingestion points: feature_description argument; content of local markdown files in docs/brainstorms/.
Boundary markers: The feature_description is wrapped in XML-like tags, but the content from brainstorm files is read and processed without explicit boundary isolation.
Capability inventory: The skill can execute shell commands (ls, mkdir, gh, linear, open, cat) and trigger other specialized research agents.
Sanitization: There is no evidence of sanitization for the content of the description or title before it is passed to research tasks or used in GitHub/Linear CLI commands beyond basic filename formatting.

workflows-plan