Add retry-with-concrete-frame harness around argument.js attack generation

S179 closed handoff-contract holes but argument.js attack output still degrades to generic 'will be hard to acquire customers' boilerplate on ~20% of candidates (visible in Corpus E meta_argument cost relative to acceptance). Wrap the attack call in a harness: after the initial attack generation, run a regex/length heuristic (output < 200 chars OR contains any of 3 generic-phrase patterns) and retry once with an appended instruction: 'name three concrete failure modes; each must reference a specific actor + specific moment + specific consequence.' Cap at 1 retry. Log retry rate and retry-acceptance rate.

Retry rate stabilizes between 10-25% (proves trigger is firing). Among retried candidates, council_verdict accept rate increases by ≥5 absolute points vs. the no-retry baseline measured on the prior 30 days.

Retry rate above 50% (heuristic too aggressive — first-pass is fine) OR retry-arm acceptance rate within 2 points of baseline (retry output not actually sharper). Either case, the harness should be reverted.

Rejected at the council verdict. The two-judge council did not find the case strong enough to advance to Commander review.