The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests user-provided project specifications to drive architectural decisions and file generation.
Ingestion points: The agent reads the user's initial project description or specification during the "Step One" evaluation phase to calculate architectural scores.
Boundary markers: While the skill sets strong internal constraints (e.g., allowMonolith: false), it lacks explicit instructions to sanitize or ignore instructions embedded within the user specification that might attempt to override these architectural rules via prompt injection.
Capability inventory: The skill utilizes Write, Edit, Bash, and Glob tools, enabling it to create complex directory structures and configuration files across multiple languages (Go, Rust, Python, TypeScript, Elixir) based on the input interpretation.
Sanitization: A deterministic Python script (scripts/evaluate-architecture.py) is used to score the specification via regex, providing a layer of objective analysis, though the final implementation plan and subagent delegation are interpreted by the agent from the raw input.

fable-architecture