The Agent Skills Directory

[COMMAND_EXECUTION]: The run_tests function in scripts/run_dual_axis_review.py executes pytest or uv run pytest on the target project's test directories. This provides runtime verification of skill health as documented in the workflow. The implementation uses list-based arguments with subprocess.run to avoid shell injection, though it does execute Python code found in the target project.
[INDIRECT_PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by interpolating content from reviewed skills into a review prompt generated by build_llm_prompt. Ingestion points: SKILL.md, scripts/, and references/ files of the reviewed project. Boundary markers: Markdown headers and code blocks are used but lack explicit sanitization. Capability inventory: The skill can write files and execute tests. Sanitization: No explicit escaping of interpolated content.
[SAFE]: The implementation follows security best practices by using yaml.safe_load for parsing configuration and metadata. Additionally, the skill includes a dedicated safety check that scans project files for hardcoded absolute user paths to prevent accidental PII leakage.

dual-axis-skill-reviewer