C2: argus methodology improvements (argus-review repo)

[SSFSKIM]

Problem & intent

Land C1's adopted candidates in the argus-review repo (~/Developer/GitHub/codex-review, remote SSFSKIM/argus-review) — the methodology upgrade that justifies making argus the default engine. Child C2 of docs/doperpowers/specs/2026-07-26-claude-review-stack-roadmap.md.

Constraints

• Changes trace one-to-one to C1 adoption decisions — no freelance methodology edits.
• Skill/agent-text changes follow the writing-skills discipline.
• Work happens in the argus-review repo; this ticket tracks it on the doperpowers board.

Success criteria

• An argus-review release whose changes trace to C1's adopt list.
• X1 benchmark re-run: new argus ≥ baseline argus, AND the argus-vs-codex delta is reported (informational early warning for the C4 gate — a large remaining gap is flagged to the human here, not discovered at C4).

Open questions

• (residue of C1's findings — grill only what C1 leaves open)

Decision log

• C2 kept separate from C3 (auto-routing): independently verifiable deliverables (roadmap Decision Log).

[SSFSKIM]

[board] ready-for-agent: roadmap C2 — blocked-by C1