The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to run various local Python scripts to generate patent screening platforms and static reports.
Evidence: SKILL.md contains multiple workflows involving python generic-drug-scout-v1\scripts\create_seeded_screening_platform.py and python .\仿药速探V1_首轮结果平台\app\server.py.
[COMMAND_EXECUTION]: The skill uses system commands or Python subprocesses to automatically launch the generated reports in the user's default browser.
Evidence: SKILL.md and scripts/create_seeded_screening_platform.py utilize Start-Process (PowerShell) and subprocess.Popen(["cmd", "/c", "start", "", html_path]) to open files.
[DYNAMIC_EXECUTION]: The script scripts/create_seeded_screening_platform.py dynamically loads and executes the server.py module from the skill's assets folder at runtime.
Evidence: The load_server_module function uses importlib.util.spec_from_file_location and spec.loader.exec_module to load code from assets/platform-template/app/server.py.
[DATA_EXFILTRATION]: The skill accesses a local configuration file to retrieve endpoints for internal Patsnap MCP services.
Evidence: assets/platform-template/app/server.py reads from ~/.codex/config.toml using the read_mcp_url function to obtain URLs for zhihuiya_logic_096456 and zhihuiya_logic_2b0355.
[INDIRECT_PROMPT_INJECTION]: The skill processes patent descriptions and legal metadata from external MCP patent tools, which constitutes an indirect prompt injection surface.
Ingestion points: External patent data is ingested via mcp_call in server.py from tools such as ls_patent_search and patsnap_fetch.
Boundary markers: The generated HTML report in index.html uses an escapeHtml JavaScript function to sanitize data before rendering.
Capability inventory: The skill can perform file system writes (creating reports), execute local Python code, and initiate network requests to MCP services.
Sanitization: Basic HTML entity escaping is implemented for strings displayed in the report UI.

generic-drug-scout