The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes shell commands to manage the installation and lifecycle of the Pyright language server.
Evidence: scripts/lsp_client.py uses subprocess.run to install the tool and subprocess.Popen to start the language server in a sidecar process using stdio for communication.
[EXTERNAL_DOWNLOADS]: The skill's setup process downloads the Pyright package from standard registries using official tools.
Evidence: scripts/bootstrap.sh and scripts/lsp_client.py automate the installation of the pyright package from well-known registries via uv or pipx.
[PROMPT_INJECTION]: The skill processes project code which creates a surface for indirect prompt injection if the analyzed files contain malicious instructions in docstrings or comments.
Ingestion points: The did_open method in scripts/lsp_client.py reads local file content and sends it to the language server.
Boundary markers: The results of semantic queries (like hover text or diagnostics) are returned as strings without explicit injection delimiters.
Capability inventory: The skill can read local files and execute the language server subprocess.
Sanitization: Output from the LSP server (e.g., inferred types, docstrings) is returned directly without additional filtering.

python-lsp