The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill instructions mandate a workflow where the agent must generate a Python script (review_pipeline.py) and execute it using the python.run tool. While this is used for report generation and data visualization, generating and running code at runtime is a high-complexity behavior that increases the potential attack surface if the generation logic is compromised.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and parse untrusted data from user-provided feasibility reports and project guidelines to extract technical routes and innovation points. This data is then used to construct search queries and populate an HTML report.
Ingestion points: SKILL.md (Step 1) uses the files tool to read external PDF or Word documents provided by the user.
Boundary markers: The instructions lack explicit boundary markers or directives to ensure the agent ignores instructions embedded within the user-provided reports during the extraction phase.
Capability inventory: The skill possesses significant capabilities, including local file writing, Python code execution (python.run), and network access via patent and web search tools.
Sanitization: There are no instructions for sanitizing or validating the content extracted from the reports before it is interpolated into the Python script or the final HTML report.
[COMMAND_EXECUTION]: The export_json.py script and the skill's execution flow involve file system operations, including reading environment variables and writing serialized JSON data to specific project directories (../feasibility-platform/data).

feasibility-review