The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's primary function involves ingesting and analyzing untrusted external data, including repository source code, design documents, and project mappings. This creates a vulnerability surface where malicious instructions embedded in a project's files could influence the agent's logic during the 'Grounding' or 'Sketching' phases.
Ingestion points: The agent is instructed to read source code, project mappings, tests, remotes, and external design documents in SKILL.md (Phase A: Ground the problem).
Boundary markers: The skill notes that 'Repository instructions and the current source outrank this skill', which establishes an authority hierarchy but does not provide a security boundary against malicious content within those files.
Capability inventory: The skill is authorized to perform file modifications in Roblox Studio fallback mode and execute rojo build commands.
Sanitization: There are no explicit instructions for the agent to sanitize or ignore potentially malicious instructions found within the project files it analyzes.
[COMMAND_EXECUTION]: The skill uses rojo build as a standard part of the Roblox development workflow to validate project assembly.
Evidence: SKILL.md Phase D specifies the use of rojo build for projects using the Rojo toolchain to validate assembly and serialization.

architect