The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the Bash tool to execute dynamically generated prototype code and collect metrics (Phase 4). While scoped to a prototype directory, this allows for the execution of any code the agent generates based on user input.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes untrusted user input to guide script generation and execution with 'intentionally relaxed' standards.
Ingestion points: User-supplied concept-description argument (Phase 1).
Boundary markers: None; the skill does not use specific delimiters or instructions to ignore embedded commands within the concept description.
Capability inventory: The skill has access to Bash, Write, Edit, and Task tools, enabling it to write files, execute shell commands, and spawn other agents.
Sanitization: No evidence of input validation or sanitization before the description is used to influence code implementation.
[COMMAND_EXECUTION]: The skill uses user input (concept-name) directly in file system paths (prototypes/[concept-name]/). Without proper sanitization by the agent, this could lead to path traversal attempts, although the isolation: worktree setting provides some environment restriction.

prototype