The Agent Skills Directory

[DYNAMIC_EXECUTION]: The scripts scripts/jax_benchmark_harness.py, scripts/jax_compile_probe.py, and scripts/jax_recompile_explorer.py implement dynamic loading of Python modules and files using importlib.import_module and exec_module based on command-line arguments. This is used to load user code for benchmarking and analysis purposes.\n- [INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection via scripts/jax_project_scan.py.\n
Ingestion points: Reads and parses local Python (.py) files provided by the user or project directory.\n
Boundary markers: The script provides structured (JSON) or unstructured (text) reports containing code snippets without using specific delimiters or 'ignore' instructions to isolate untrusted content from the agent's prompt context.\n
Capability inventory: The skill possesses capabilities for command execution, file system access, and dynamic Python code loading/execution via bundled scripts.\n
Sanitization: The script performs AST-based parsing but does not sanitize or filter the content of extracted code snippets or comments before they are presented to the agent.

jax-development