kf-g-agent-review-adversarial-after-primary-response

Adversarial verification after the initial answer

After finishing the design or implementation and giving the user an initial answer, adversarially verify the deliverable and connect the findings to self-improvement.

Minimum execution order

1. Confirm that the initial answer has been given. If not, give the initial answer first → procedures.md
2. Launch verification according to the type of deliverable → procedures.md
3. Confirm the boundaries with other skills → boundaries.md
4. Complete handling of the findings and the addendum → procedures.md

references table of contents

File	Reference start conditions
procedures.md	Confirmation of the initial answer, launching verification, countermeasures, addendum
boundaries.md	Boundaries with other skills, situations where this does not apply

Final check

1. The initial answer is given before verification
2. For design, kf-g-agent-review-adversarial-evaluation is followed
3. For implementation, use the verification worker and kf-g-agent-review-post-implementation-two-stage , and do not substitute the adversarial evaluation skill
4. There is a countermeasure for each finding. If there are substantive branches, list the options, and put the recommendation at the end of the line (推奨)(for HTML, a badge). If there is essentially only one choice, state that it has been adopted/reflected