The Agent Skills Directory

[SAFE]: The skill is designed exclusively for static test design and metamorphic analysis. It contains explicit constraints in SKILL.md and prompts/metamorphic-testing.md that prohibit the agent from executing tests, editing code or requirements, and claiming test passes or coverage.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an ingestion surface for indirect prompt injection as it is designed to analyze external material such as requirements, designs, and code fixtures. However, this risk is mitigated by the skill's strict 'no-execution' policy and the requirement to maintain a 'six-part input audit' which forces the agent to explicitly identify assumptions and missing information before proceeding with analysis.
[COMMAND_EXECUTION]: There is no evidence of arbitrary command execution. The skill instructions focus on generating markdown reports and tables rather than invoking system tools or scripts.
[METADATA_POISONING]: The metadata in SKILL.md and agents/openai.yaml is consistent with the skill's stated purpose of metamorphic test design.

metamorphic-testing