The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it ingests untrusted data from project-level files.
Ingestion points: Phase 1 (Discovery) reads content from .forgewright/product-manager/BRD/brd.md and .forgewright/polymath/handoff/context-package.md to derive architecture constraints.
Boundary markers: No specific boundary markers or instructions to ignore embedded commands are used when interpolating this external data into the agent context.
Capability inventory: The skill possesses the capability to scaffold an entire project structure, including creating directories, writing source code, and generating Dockerfiles.
Sanitization: There is no evidence of validation or sanitization of the input data from the BRD or handoff documents before they are used to guide the architecture design process.
[COMMAND_EXECUTION]: The skill utilizes dynamic context injection to load shared protocols and local configuration.
Evidence: Multiple !cat commands are used in the 'Protocols' and 'Engagement Mode' sections (e.g., !cat .production-grade.yaml).
Analysis: These commands are restricted to specific, relative paths within the skill or project directory. They do not accept user-supplied arguments or perform network operations, and are used for legitimate initialization of the skill's operating environment.

solution-architect