The Agent Skills Directory

EXTERNAL_DOWNLOADS (MEDIUM): The documentation in QUICKSTART.md and CHANGELOG.md instructs users to install the pypict Python library via pip install pypict.
Evidence: This is a third-party dependency from an individual maintainer (Kenichi Maehashi) rather than a trusted organization. While legitimate for its purpose, it constitutes an external code dependency that must be verified.
COMMAND_EXECUTION (MEDIUM): The STRUCTURE.md file identifies a Python utility script scripts/pict_helper.py designed for PICT generation and formatting.
Evidence: The skill's workflow necessitates the execution of this local script or the pypict library by the agent to fulfill its core function of generating test cases. Execution of scripts bundled within a skill is a sensitive operation.
PROMPT_INJECTION (MEDIUM): The skill is designed to process external, potentially untrusted content (system specifications and user requirements) to generate its output.
Evidence: As described in QUICKSTART.md, the skill ingests user requirements to identify parameters and write constraints. Maliciously crafted requirements could use indirect prompt injection techniques to manipulate the agent's logic during the test design process.
Ingestion Points: examples/atm-specification.md and user-provided natural language requirements.
Capability Inventory: Local script execution (pict_helper.py) and result formatting.
Boundary Markers: None identified in the provided documentation to distinguish untrusted requirement data from internal reasoning.
Sanitization: No evidence of sanitization or validation of the input requirements before processing.

pict-test-designer