The Agent Skills Directory

[SAFE]: The skill facilitates manual QA workflows by reading local plan files and writing results. It operates entirely on the local file system and via browser interaction.
[SAFE]: There are strong negative constraints that prevent the agent from using administrative tools (terminal, DB, GraphQL), ensuring it stays within the 'audience constraints' of the plan and does not perform unauthorized system actions.
[SAFE]: Browser automation usage is restricted to actions a human would perform, with explicit rules against bypassing constraints via scripts or shortcuts like the JS console or CDP evaluate().

qa-run