The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill recommends using pickle and joblib for serializing Python objects to disk for caching and data loading purposes.\n
The references/caching.md file advises that "Values are pickled to disk" when using diskcache for cross-process sharing.\n
The references/data-loading.md file suggests using pickle or joblib for non-tabular artifacts to avoid text format overhead.\n
pickle is unsafe for deserializing data from untrusted sources as it can lead to arbitrary code execution.\n- [INDIRECT_PROMPT_INJECTION]: The skill's primary function is to analyze project source code, creating a surface for malicious instructions to be processed by the agent.\n
Ingestion points: The agent reads app.py and other modules in the application project.\n
Boundary markers: No specific delimiters or warnings are used to mark the ingested code as untrusted.\n
Capability inventory: The skill includes capabilities to write files (outputs/measurements/) and execute shell commands for application profiling and tracing.\n
Sanitization: There is no evidence of sanitization or filtering of the ingested source code content.\n- [EXTERNAL_DOWNLOADS]: The skill references and encourages the installation of external tools for performance analysis.\n
Fetches the shinyloadtest tool from the NPM registry, which is an official tool from the vendor (posit-dev).\n
Recommends installing the py-spy sampling profiler and shiny[otel] via pip, which are well-known diagnostic tools.\n- [COMMAND_EXECUTION]: The skill provides numerous shell commands for the agent or user to execute for diagnostic purposes.\n
Commands include running the application with OpenTelemetry tracing (opentelemetry-instrument) and using py-spy for CPU profiling.

shiny-py-optimize