The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze numerous project files, including route handlers, service logic, and configuration files like pyproject.toml. These files are considered untrusted external data. If an attacker embeds malicious instructions within these files (e.g., in code comments or string literals), the agent might execute those instructions while performing its audit.
Ingestion points: The skill instructions in SKILL.md specifically direct the agent to "Read each route file", "Read each service file", "Read model files", "Read util files", "Read agent files", "Read main.py", and "spot-check pyproject.toml files".
Boundary markers: The instructions lack explicit boundary markers or warnings to the agent to disregard natural language instructions found within the source code files being analyzed.
Capability inventory: The agent uses file system read capabilities across the entire project structure and generates a detailed report based on the content of these files.
Sanitization: No sanitization or filtering of the file content is mentioned before the agent processes it for rule violations.
[EXTERNAL_DOWNLOADS]: The skill specifies a dependency on an external instruction set that is not contained within the skill package itself. It instructs the agent to load and follow rules from a separate skill named fastapi-rules.
Evidence: SKILL.md states: "The authoritative rules... live in the fastapi-rules skill. Before auditing, load the fastapi-rules skill and treat every rule there as the checklist for this review."

fastapi-architect