The Agent Skills Directory

[DYNAMIC_EXECUTION]: The scripts/generate_test_suite.py script includes a generate_script method that assembles a Bash script using a string template provided via the --content argument and writes it to the local file system. This allows for the creation of arbitrary executable content at runtime.
[PRIVILEGE_ESCALATION]: The generate_script method in scripts/generate_test_suite.py uses file_path.chmod(0o755) to programmatically grant execution permissions to the scripts it generates. While intended for automation, this pattern is used to escalate the capabilities of dynamically created files.
[COMMAND_EXECUTION]: The skill's primary purpose involves generating and running test suites across multiple environments (Node.js, Python, Java). The YAML frontmatter explicitly allows the Bash tool, and the instructions direct the agent to generate and potentially execute these scripts, posing a risk if the generated content is malicious.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and parse untrusted external data, specifically OpenAPI and GraphQL specifications.
Ingestion points: The skill reads API specification files (e.g., assets/example_openapi.yaml, assets/example_graphql_schema.graphql) to catalog endpoints and schemas.
Boundary markers: There are no boundary markers or specific instructions to treat embedded strings within the specifications as untrusted data.
Capability inventory: The skill possesses file writing, permission modification (chmod), and bash execution capabilities.
Sanitization: The generate_test_suite.py script performs no sanitization or escaping of the content or template strings before embedding them into executable bash scripts, making it vulnerable to command injection if the input specifications contain malicious payloads.

automating-api-testing