The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill implements a workflow that ingests untrusted data from external sources to influence agent behavior during design evaluation.
Ingestion points: Processes content from a user-provided BRIEF document and a copy.json file in SKILL.md.
Boundary markers: Employs logical constraints such as the "Sprint Contract" to limit scoring criteria, but lacks explicit content delimiters or instructions to ignore embedded commands within the ingested data.
Capability inventory: The skill is capable of writing artifacts to the .moai/sprints directory and executing shell-based testing commands.
Sanitization: No evidence of sanitization, escaping, or validation of the BRIEF content is present in the implementation guide.
[DYNAMIC_EXECUTION]: The skill generates and invokes shell commands for automated design verification.
Evidence: The sprint-N.json structure defines a test_scenarios field used to execute tools like Playwright (e.g., playwright test --viewport 375x667). This represents script or command generation from a predefined template.

design-iteration-loop