The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill parses external, potentially untrusted API definition files (OpenAPI, Postman, WSDL, curl) using scripts/parse_api_sources.py to drive test generation.
Ingestion points: Untrusted data enters the agent context through multiple file formats parsed by scripts/parse_api_sources.py (e.g., _parse_openapi, _parse_postman, _parse_wsdl).
Boundary markers: While prompts/api-test-restassure.md provides instructions for the agent to be faithful to the input, it lacks explicit boundary markers or sanitization logic to prevent malicious instructions embedded in API descriptions from influencing the agent's code generation logic.
Capability inventory: The skill has the capability to write files (scripts/generate_restassured_tests.py) and execute them via shell commands (scripts/run.sh calling mvn test).
Sanitization: The skill relies on standard parsers like json and yaml.safe_load, which do not sanitize against natural language prompt injection attacks.
[DYNAMIC_EXECUTION]: The skill implements a workflow that generates Java source code (GeneratedApiTest.java) based on external inputs and subsequently executes it using Maven.
Evidence: scripts/generate_restassured_tests.py builds the Java class string, and scripts/run.sh executes mvn -q test on the result.
Risk: Maliciously crafted API definitions could potentially influence the generated code to perform unauthorized actions during the execution phase.
[DYNAMIC_EXECUTION]: The WSDL parsing logic in scripts/parse_api_sources.py uses an insecure XML parsing method.
Evidence: The _parse_wsdl function uses xml.etree.ElementTree.fromstring(text) to process WSDL files.
Risk: Standard Python xml.etree.ElementTree is vulnerable to XML External Entity (XXE) attacks. An attacker providing a malicious WSDL file could potentially read local files or conduct Server-Side Request Forgery (SSRF) during the parsing phase.
[COMMAND_EXECUTION]: The README.md file encourages users to run installation scripts with relaxed security constraints on Windows.
Evidence: The installation instructions for Windows use powershell -ExecutionPolicy Bypass.
Risk: While common for development tools, bypassing execution policies allows the execution of unsigned scripts, which increases the risk if the local environment or the installation scripts were compromised.

api-test-restassure