The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external OpenAPI specifications and SQL schema files. Maliciously crafted inputs could potentially influence the generated code or analysis results.
Ingestion points: scripts/api_scaffolder.py (OpenAPI specs), scripts/database_migration_tool.py (SQL files).
Boundary markers: Absent. The tools process these files directly without checking for embedded instructions.
Capability inventory: scripts/api_scaffolder.py (file-write), scripts/database_migration_tool.py (file-write), scripts/api_load_tester.py (network ops).
Sanitization: Basic regex is used for naming conventions in scripts/api_scaffolder.py, but core content from schema definitions is interpolated into code templates.
[DYNAMIC_EXECUTION]: The scripts/api_scaffolder.py script dynamically generates TypeScript route handlers, validation schemas, and types based on provided OpenAPI specifications.
Evidence: generate_routes, generate_validators, and generate_types methods in scripts/api_scaffolder.py assemble source code strings and write them to the file system.
[DATA_EXFILTRATION]: The scripts/api_load_tester.py script provides the ability to perform network requests to arbitrary endpoints with custom headers and request bodies. While intended for testing, this capability could be misused to probe internal networks or exfiltrate data. The tool also includes a --no-verify-ssl flag that disables SSL certificate verification.
[COMMAND_EXECUTION]: The skill instructs the agent to execute Python scripts (api_scaffolder.py, database_migration_tool.py, api_load_tester.py) that perform file system and network operations. These tools take user-provided arguments (filenames, connection strings, URLs) which could lead to command injection if not properly handled by the agent context.

senior-backend