The Agent Skills Directory

[DYNAMIC_EXECUTION]: The scripts/api_scaffolder.py script generates Express.js route handlers by directly interpolating paths and methods from an OpenAPI specification into string templates. Because these values are not properly escaped, a malicious specification can include sequences (e.g., closing quotes and semicolons) to inject and execute arbitrary code within the generated files. This poses a high risk if an agent or developer scaffolds code from an untrusted specification and then executes the resulting application.
[INDIRECT_PROMPT_INJECTION]: The skill provides tools (api_scaffolder.py and database_migration_tool.py) that ingest potentially untrusted external files such as OpenAPI specs and SQL schemas. These ingestion points allow for indirect prompt injection, where malicious instructions or payloads within the data files can influence the generated code or system analysis without direct user interaction.
Ingestion points: scripts/api_scaffolder.py (OpenAPI specs) and scripts/database_migration_tool.py (SQL schema files).
Boundary markers: The script templates do not use boundary markers or delimiters to isolate untrusted input from the generated code structure.
Capability inventory: api_scaffolder.py has file-writing capabilities; api_load_tester.py can perform network requests.
Sanitization: Sanitization is limited to basic regex for identifiers and does not prevent code injection in URL paths or other schema-derived fields.

senior-backend