feat: remove behavior diff run mode prompt

[kentwelcome]

Summary

• Start Behavior Diff as soon as the comparison task is known, without asking about run modes, counts, or model cost.
• Keep --fast available only when the user explicitly asks for it.
• Carry the exact demo fixture task into the initial prompt, keep both host manifests aligned at 0.3.2, and enforce the behavior with deterministic contracts.

Test plan

• docker run --rm -v "$PWD:/mnt" -w /mnt mvdan/shfmt:v3.14.0 -d -i 2 -ci .
• uvx ruff@0.16.5 format --check --diff .
• bash -n bin/behavior-diff plugin/scripts/*.sh plugin/skills/behavior-diff/scripts/*.sh tests/*.sh
• python3 -m py_compile plugin/skills/behavior-diff/scripts/decisions.py plugin/skills/behavior-diff/scripts/render.py
• bash tests/hooks-test.sh
• python3 plugin/skills/behavior-diff/scripts/decisions.py --check
• bash tests/live-report-contract.sh
• claude plugin validate plugin
• git diff --check

Generated with OpenAI Codex

[Copilot]

🟢 Approval recommended

The updated skill instructions, docs, and deterministic contracts are internally consistent and the changes are low-risk and well-enforced by the added contract checks.

Pull request overview

This PR updates the Behavior Diff “headless” skill and the demo/nudge harness documentation/contracts so the comparison run starts immediately once the task is known, without prompting the user for run modes, trial counts, or model cost, while still supporting --fast only on explicit request.

Changes:

• Update plugin/skills/behavior-diff/SKILL.md to remove the “confirm before running” gate and require immediate execution with host preservation via --agent <current-host>.
• Strengthen deterministic contracts in tests/live-report-contract.sh to enforce the new wording/behavior expectations (including carrying the exact fixture task into the initial prompt).
• Refresh demo and e2e guidance (tests/nudge-e2e.sh, .agents/.../SKILL.md, e2e/README.md) to match the new flow and remove references to cost/run-mode gating.

File summaries

File	Description
tests/nudge-e2e.sh	Includes the exact fixture task as a “later request” in the setup prompt and updates acceptance instructions to pass that exact task into behavior-diff.
tests/live-report-contract.sh	Adds deterministic string/line contracts to enforce immediate start behavior, --fast gating, host preservation, and fixture-task carry-through; bumps version checks to 0.3.2.
plugin/skills/behavior-diff/SKILL.md	Replaces “confirm before running” with “run immediately once task is known,” removes cost/mode mentions, and documents --agent <current-host> usage.
plugin/.codex-plugin/plugin.json	Bumps plugin version to 0.3.2 to stay aligned with Claude manifest.
plugin/.claude-plugin/plugin.json	Bumps plugin version to 0.3.2 to stay aligned with Codex manifest.
e2e/README.md	Removes references to cost/confirmation gating and clarifies correctness requirements around reusing the exact task.
.agents/skills/run-behavior-diff-demo-journey/SKILL.md	Updates the live demo steps to accept the ask by supplying the exact task and removes cost/--fast prompting guidance.

Review details

• Files reviewed: 7/7 changed files
• Comments generated: 0
• Review effort level: Lite

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.