The Agent Skills Directory

[SAFE]: The skill serves as a quality-control framework for financial alpha research, focusing on identifying logical errors such as data leakage and unrealistic cost assumptions.
[INDIRECT_PROMPT_INJECTION]: The skill ingests user-supplied factor definitions, creating a theoretical attack surface. However, the instructions proactively mitigate this by requiring the agent to adopt a 'suspect-handling' stance and rewrite all inputs into an auditable format before analysis, establishing strong cognitive boundaries.
[Ingestion points]: SKILL.md (factor_definition parameter).
[Boundary markers]: The workflow mandates a 'suspect-handling' (嫌疑人处理) posture and auditable language translation.
[Capability inventory]: The skill is restricted to text-based analysis and contains no file-writing, network, or shell-execution capabilities.
[Sanitization]: The skill relies on cognitive restructuring of input rather than programmatic filtering.
[COMMAND_EXECUTION]: The project includes a validation script (scripts/check_test_cases.py) used for internal documentation checks. The script uses only standard Python libraries and does not accept untrusted network input, representing standard development practice.
[EXTERNAL_DOWNLOADS]: All external URLs point to standard metadata badges (shields.io) or the author's own GitHub repository (quantskills), which is considered a legitimate vendor resource.

factor-mason