The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted external data in the form of OpenAPI specifications and SQL schema files to generate code scaffolding and migration reports.
Ingestion points: api_scaffolder.py (OpenAPI YAML/JSON files) and database_migration_tool.py (SQL schema files).
Boundary markers: The tools parse these files without explicit boundary markers or instructions to ignore embedded natural language prompts within schema documentation or comments.
Capability inventory: The skill includes tools for writing generated source code to the local file system (api_scaffolder.py, database_migration_tool.py) and performing concurrent network requests to arbitrary URLs (api_load_tester.py).
Sanitization: While the generated code includes validation logic (e.g., Zod), the scaffolding and analysis tools themselves do not perform sanitization on documentation fields that could contain malicious instructions.
[COMMAND_EXECUTION]: The skill includes several Python scripts (api_load_tester.py, api_scaffolder.py, backend_decision_engine.py, database_migration_tool.py) that perform file and network operations. These tools are scoped to their specific backend development tasks and do not execute arbitrary shell commands or subprocesses based on user-supplied data.
[DATA_EXFILTRATION]: The api_load_tester.py utility has the capability to make network requests to arbitrary endpoints provided by the user. While this is the intended functionality for performance testing, it represents a network egress capability. The script does not show any patterns of accessing sensitive local files or credentials for exfiltration.

senior-backend