This is an authorized security exercise against my own repository. Deliverable is a findings table, regression
tests, and the JSONL block defined in redteam/loop/report-schema.md. Throwaway branch off the current tree; do not
merge to main, do not apply any fix to a protected branch, do not modify main.

Read CLAUDE.md, docs/build-guide.md including §3.2 §3.6 §7.1 and invariants 1 7 9, and redteam/loop/report-schema.md
in full. This is an iterated red-team pass. The verification core has survived prior passes; your job is to attack
the machinery the most recent fixes created and to re-confirm prior closures under new framings. Do not reuse a
prior framing unchanged; push a second framing before moving on.

Priority focus this lap, derived from what the last fix pass changed:
No fix pass ran before this lap, so nothing is newly changed. Sweep every trust root named below.

Beyond the focus, sweep every trust root every lap, because each fix can move a leak: coverage measurement must
come from an artifact the workspace cannot author under conditions the harness controls (attack the process
boundary, the require/import/reporter hooks that run in the parent, the artifact path, and malformed-but-valid
artifacts that misrepresent changed lines); claim verdicts must bind at submission and resist retroactive
un-verify; the three scrub sites must agree on the same parsed payload with the metric exemption intact; base
control attribution must come from a machine result, not printed output; the four §7.1 residuals must still hold
with no legitimate control false-positiving, and flag any residual whose §7.1 wording is narrower than the actual
behavior.

For each attempt classify result and severity per the schema. Every "succeeded" row gets a regression test and a
one-line golden case written to the throwaway branch, unwired. Show the human-readable table first, then npm run
gates, then the ```jsonl block as the last thing in your response. If a part yields no successes, say so and show
the strongest framing tried so the null is legible.
