paiml/aprender#1846 closes the env-var plumbing gap noted in M288.
M288's "NOT YET shipped" caveat is now resolved end-to-end.

## End-to-end flow now wired

```
operator shell ENV
  → bench script (inherits)
    → ccpa-arena-bench (inherits)
      → apr code (inherits)
        → AprServeDriver::build_openai_body (READS env vars)
          → HTTP POST /v1/chat/completions {temperature, top_k, ...}
            → try_qwen3_moe_backend (PARSES request)
              → QuantizedGenerateConfig {...}
                → run_qwen3_moe_generate
                  → sample_from_logits (APPLIES sampling + penalty)
```

Every link wired. Operator's `APR_AGENT_TEMPERATURE=0.3` (etc) now
flows through to actual logit sampling, no longer a no-op.

## Status reconciliation

8/10 aprender M32d-arc PRs MERGED. 2 OPEN:
- #1835 (streaming SSE contract; workspace-test pending)
- #1846 (this M289's prerequisite; just opened)

## Companion-side state

CCPA M281-M288 + M289 = 9 docs tracking the full upstream arc +
dispatch recipe + plumbing confirmation.

## What's NOT done

- V1_004 sub-bench not yet dispatched (operator-coordinated; needs
  #1846 merge + apr rebuild + ~10-15hr wall)
- Currently-running greedy baseline bench should finish first; don't
  start the 3-knob bench until the baseline scores.json lands (the
  COMPARISON is the value)

Mechanical doc. M-counter NOT bumped.

Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>

…4 follow-up) (#1849)

Adds 3 concrete few-shot <tool_call> examples to CODE_SYSTEM_PROMPT
(the 7B+ branch used for Qwen3-Coder-30B-A3B). Empirical context:
paiml/claude-code-parity-apr M287 evidence showed the 30B model
emits Markdown ```rust``` code blocks (in turn-1 text) instead of
<tool_call> JSON. The parser at realizar.rs:144-149 accepts
<tool_call> + ```json``` but NOT ```rust``` — so the model's
turns are silently text-only, bench hits per-turn timeout after 4
turns of rambling.

The 3-knob toolkit (sampling/penalty/streaming) tunes probability
distributions but can't change format adherence. THIS PR addresses
the format adherence directly by:

1. Showing the model 3 concrete <tool_call> examples in-context
   (file_read, file_edit, shell)
2. Adding an explicit "ALWAYS gets a tool-call response" rule
3. Adding "Be concise — DO NOT narrate" guideline
4. Adding "DO NOT use Markdown ```rust``` code blocks" anti-rule

## Why few-shot examples work

Large language models are pattern-matchers. Showing them the exact
format they should emit (rather than just describing it) drastically
improves format adherence on coder-finetuned models. The 30B-Coder
has strong "Markdown code block" priors from training; explicit
counter-examples + the negative rule pull it toward the <tool_call>
format.

## Empirical context

M287 (Phase 6 bench, fixtures 1-10 + greedy decoding): uniform
driver_error / turns_before_error=4 pattern. Every turn was text
with Rust code in Markdown, no tool calls extracted. Operator
playbook calls for sampling/penalty sub-bench (#1842 + #1844 + #1846
shipped). This PR is COMPLEMENTARY: prompt fix + sampling together
have the best chance of breaking the rambling pattern.

## Companion-side dispatch (post-merge)

After this PR + rebuild, operator can run a NEW sub-bench (call it
Sub-bench E in M288 nomenclature) that combines:
- 3-knob sampling (temperature=0.3, top_k=50, top_p=0.95)
- Repetition penalty (repeat_penalty=1.2, repeat_last_n=64)
- THIS PR's few-shot prompt (active by default; no env var needed)

If Sub-bench E shows ANY fixture pass, V1_004 discharges.

Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>