@event4u/agent-config 1.15.0 → 1.17.0

package/scripts/ai_council/_one_off_phase_2a_budget_rebalance.py

@@ -0,0 +1,257 @@
+"""Council adjudication of Phase 2A always-budget overshoot.
+
+Phase 2A of road-to-structural-optimization slimmed the top-3
+budget-consuming `type: always` rules (language-and-tone,
+ask-when-uncertain, direct-answers) by extracting MECHANICS into
+contexts under contexts/communication/rules-always/. Under the
+locked Model (b) literal, the extended budget grew from 47,448 to
+52,534 chars (49,000 cap → 107.2 %), 5.2 pp above the 2 % G3
+tolerance band defined in
+docs/contracts/load-context-budget-model.md.
+
+Per that contract: an overshoot above the band rejects the model
+and escalates to council. Phase 2A § Abort/rollback fires its
+budget kill-switch in this exact case.
+
+This one-off bundles the linter output, the three slimmed rules,
+the three new mechanics contexts, and the locked budget contract,
+and asks the council to pick the smallest viable resolution.
+
+Invocation:
+    .venv/bin/python -m scripts.ai_council._one_off_phase_2a_budget_rebalance
+"""
+from __future__ import annotations
+
+import sys
+from pathlib import Path
+
+from scripts.ai_council.bundler import bundle_files
+from scripts.ai_council.clients import (
+    AnthropicClient,
+    OpenAIClient,
+    load_anthropic_key,
+    load_openai_key,
+)
+from scripts.ai_council.orchestrator import (
+    CostBudget,
+    CouncilQuestion,
+    consult,
+    estimate,
+)
+from scripts.ai_council.pricing import estimate_cost, load_prices
+from scripts.ai_council.project_context import detect_project_context
+from scripts.ai_council.session import SessionManifest, save as save_session
+
+REPO_ROOT = Path(__file__).resolve().parents[2]
+ARTEFACTS = [
+    REPO_ROOT / "docs/contracts/load-context-budget-model.md",
+    REPO_ROOT / ".agent-src.uncompressed/rules/language-and-tone.md",
+    REPO_ROOT / ".agent-src.uncompressed/rules/ask-when-uncertain.md",
+    REPO_ROOT / ".agent-src.uncompressed/rules/direct-answers.md",
+    REPO_ROOT / ".agent-src.uncompressed/contexts/communication/rules-always/language-and-tone-mechanics.md",
+    REPO_ROOT / ".agent-src.uncompressed/contexts/communication/rules-always/ask-when-uncertain-mechanics.md",
+    REPO_ROOT / ".agent-src.uncompressed/contexts/communication/rules-always/direct-answers-mechanics.md",
+]
+
+ORIGINAL_ASK = (
+    "Phase 2A of road-to-structural-optimization slimmed the top-3 "
+    "always-rules and breached the 2 % G3 tolerance band on the "
+    "always-budget under Model (b) literal. The contract requires "
+    "council escalation. Council task: pick the smallest viable "
+    "resolution from a fixed option set."
+)
+
+REVIEW_PROMPT = """\
+# Council Adjudication — Phase 2A Always-Budget Overshoot
+
+## Measured state (linter output, model (b) literal)
+
+```
+FAIL  always-rule extended budget: 52,534 / 49,000 chars (107.2%)
+      thresholds: warn 80% · fail 90% · per-rule ≤ 6,000 (ext) ·
+                  top-3 ≤ 24,500 (ext) · depth ≤ 2 · G3 band ≤ +2%
+
+  ext= 8529  raw= 4636  scope-control.md (top-3)  allowlisted ≤ 8,529
+  ext= 7887  raw= 4607  non-destructive-by-default.md (top-3)  allowlisted ≤ 7,887
+  ext= 6827  raw= 2424  ask-when-uncertain.md (top-3)  PER-RULE BREACH
+  ext= 6283  raw= 2758  direct-answers.md  PER-RULE BREACH
+  ext= 5863  raw= 3658  language-and-tone.md
+  ext= 5781  raw= 3309  commit-policy.md
+  ext= 5481  raw= 2196  verify-before-complete.md
+  ext= 4415  raw= 4415  no-cheap-questions.md
+  ext= 1468  raw= 1468  agent-authority.md
+```
+
+Pre-Phase-2A baseline (Phase 0.2.3 retroactive test, locked):
+**47,448 / 49,000 = 96.8 %** (within tolerance, model (b) accepted).
+
+Phase-2A delta per slimmed rule (raw rule shrank, mechanics
+context loaded as new dependency):
+
+| Rule | ext before | ext after | delta |
+|---|---:|---:|---:|
+| language-and-tone | 5,832 | 5,863 | +31 |
+| ask-when-uncertain | 5,196 | 6,827 | +1,631 |
+| direct-answers | 4,722 | 6,283 | +1,561 |
+
+Phase 0.4 worked example council-locked the +1,561 delta on
+direct-answers. ask-when-uncertain (+1,631) is structurally
+identical. The total overshoot is **+3,534 chars over cap**.
+
+## Why the literal model can no longer hit budget
+
+Under model (b) literal each rule pays full cost for every context
+it loads. Each mechanics extraction adds ~1.5–2 KB of frontmatter +
+headers + context-introduction overhead per rule. Phase 2A is
+structurally unable to satisfy the roadmap success criterion
+("budget delta ≥ −5 %") under model (b); extraction *always* adds
+overhead unless mechanics are consolidated and shared across rules.
+
+## Fixed option set (pick exactly one)
+
+- **R1 — Switch to Model (c) shared-divisor.** Reserved by the
+  contract as "first refinement step if the 2 % tolerance band is
+  exceeded". Under (c), `Σ RawSize(c) / N_loaders`. Currently each
+  mechanics file has N=1, so (c) reduces to (b). Buy-in for (c)
+  enables R5 below.
+- **R2 — Raise TOTAL_CAP.** Set cap to 56,000 chars (current
+  utilization 93.8 % under new cap). Acknowledges extraction
+  overhead is real; freezes Phase 2A wins; abandons the −5 %
+  delta success criterion.
+- **R3 — Raise PER_RULE_CAP to 7,000 + add allowlist entries +
+  raise TOTAL_CAP to 54,000.** Minimal-change variant: keeps the
+  shape, accepts the two new per-rule breaches, raises cap
+  modestly.
+- **R4 — Revert Phase 2A.** Abandon mechanics extraction on the
+  three rules; restore them to original size; mark Phase 2A as
+  "structurally infeasible under locked contract"; close roadmap
+  phase.
+- **R5 — Consolidate the three mechanics into one shared context
+  + adopt R1.** Single `rules-always-mechanics.md` loaded by all
+  three slimmed rules; under (c) each rule pays 1/3 of the
+  consolidated context size. Highest engineering cost, only path
+  that delivers a net negative delta.
+
+## Output Contract (STRICT)
+
+```
+### Verdict
+**Choice:** <R1 | R2 | R3 | R4 | R5>
+**One-sentence rationale:** <≤ 30 words>
+```
+
+```
+### Required follow-up actions (numbered, ≤ 4)
+  1. <smallest concrete step, files named>
+  2. <...>
+```
+
+```
+### Risk note
+**Single biggest risk of the chosen option:** <one sentence>
+**Mitigation:** <one sentence>
+```
+
+```
+### Contract amendment needed?
+**Amend load-context-budget-model.md?** <YES — section · NO>
+**Amend road-to-structural-optimization Phase 2A success criterion?**
+  <YES — new criterion in 1 line · NO>
+```
+
+Be decisive — total response ≤ 1,000 words. Artefacts follow
+verbatim.
+"""
+
+
+def main() -> int:
+    anthropic = AnthropicClient(api_key=load_anthropic_key(), model="claude-sonnet-4-5")
+    openai = OpenAIClient(api_key=load_openai_key(), model="gpt-4o")
+    members = [anthropic, openai]
+
+    context = bundle_files(ARTEFACTS)
+    project = detect_project_context(REPO_ROOT)
+    table = load_prices()
+
+    user_prompt = REVIEW_PROMPT + "\n\n---\n\n" + context.text
+
+    question = CouncilQuestion(
+        mode="files",
+        user_prompt=user_prompt,
+        max_tokens=2560,
+    )
+
+    estimates = estimate(
+        question, members, table, project=project, original_ask=ORIGINAL_ASK,
+    )
+    print("=== ESTIMATE (single round) ===")
+    total_est = 0.0
+    for c, e in zip(members, estimates):
+        print(f"  {c.name}/{c.model}: ~{e.input_tokens} in + {e.output_tokens} out = ${e.total_usd:.4f}")
+        total_est += e.total_usd
+    print(f"  TOTAL per round (max): ${total_est:.4f}")
+    print()
+
+    budget = CostBudget(
+        max_input_tokens=200_000,
+        max_output_tokens=80_000,
+        max_calls=20,
+        max_total_usd=2.50,
+    )
+
+    rounds_collected: list[list] = []
+
+    def _on_round_complete(round_idx: int, round_responses) -> None:
+        rounds_collected.append(list(round_responses))
+        print(f"=== ROUND {round_idx + 1} COMPLETE ===")
+        for r in round_responses:
+            if r.error:
+                print(f"  [error] {r.provider}/{r.model}: {r.error}")
+                continue
+            actual = estimate_cost(r.provider, r.model, r.input_tokens, r.output_tokens, table)
+            print(f"  [done] {r.provider}/{r.model}: {r.input_tokens} in / "
+                  f"{r.output_tokens} out · {r.latency_ms} ms · ${actual.total_usd:.4f}")
+        print()
+
+    print("=== CONSULT (1 round, Phase 2A budget rebalance) ===")
+    consult(
+        members, question, budget,
+        rounds=1,
+        on_round_complete=_on_round_complete,
+        table=table, project=project, original_ask=ORIGINAL_ASK,
+    )
+
+    if not rounds_collected:
+        print("[error] no rounds completed", file=sys.stderr)
+        return 1
+
+    actual_total = 0.0
+    for round_responses in rounds_collected:
+        for r in round_responses:
+            if r.error:
+                continue
+            actual = estimate_cost(r.provider, r.model, r.input_tokens, r.output_tokens, table)
+            actual_total += actual.total_usd
+    print(f"=== TOTAL ACTUAL: ${actual_total:.4f} ===")
+
+    final_round = rounds_collected[-1]
+    if not [r for r in final_round if not r.error]:
+        return 1
+
+    manifest = SessionManifest(
+        mode="files",
+        artefact="agents/roadmaps/road-to-structural-optimization.md",
+        original_ask=ORIGINAL_ASK,
+        members=[f"{r.provider}/{r.model}" for r in final_round],
+        rounds=len(rounds_collected),
+        cost_usd_estimated=total_est,
+        cost_usd_actual=actual_total,
+        extra={"purpose": "Phase 2A always-budget overshoot adjudication"},
+    )
+    session_dir = save_session(manifest=manifest, responses=rounds_collected)
+    print(f"[saved] {session_dir.relative_to(REPO_ROOT)}/")
+    return 1 if any(r.error for round_r in rounds_collected for r in round_r) else 0
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())

package/scripts/ai_council/_one_off_phase_2a_post_revert.py

@@ -0,0 +1,197 @@
+"""Council adjudication of post-revert always-budget contract.
+
+Phase 2A executed R4 (revert) per the prior council verdict. The
+three slimmed always-rules and their mechanics contexts have been
+restored; mechanics-context directory removed. Measured state on
+this branch is now 47,448 / 49,000 chars = 96.8 % — strictly better
+than `main` at 100.6 % which currently passes via the G3 tolerance
+band.
+
+The strict 90 % FAIL gate in scripts/check_always_budget.py was
+written assuming Phase 2A would succeed in dropping utilization
+below 90 %. Phase 2A is now structurally infeasible under model (b)
+(per the previous council finding) and the cap-tightening was
+deferred to Phase 5. Branch must unblock CI without regressing the
+contract spirit.
+
+Council task: pick the smallest viable reconciliation.
+"""
+from __future__ import annotations
+
+import sys
+from pathlib import Path
+
+from scripts.ai_council.bundler import bundle_files
+from scripts.ai_council.clients import (
+    AnthropicClient,
+    OpenAIClient,
+    load_anthropic_key,
+    load_openai_key,
+)
+from scripts.ai_council.orchestrator import (
+    CostBudget,
+    CouncilQuestion,
+    consult,
+    estimate,
+)
+from scripts.ai_council.pricing import estimate_cost, load_prices
+from scripts.ai_council.project_context import detect_project_context
+from scripts.ai_council.session import SessionManifest, save as save_session
+
+REPO_ROOT = Path(__file__).resolve().parents[2]
+ARTEFACTS = [
+    REPO_ROOT / "docs/contracts/load-context-budget-model.md",
+    REPO_ROOT / "scripts/check_always_budget.py",
+]
+
+ORIGINAL_ASK = (
+    "Phase 2A reverted (R4) per prior council verdict. Branch budget is "
+    "96.8 % (gap-zone FAIL); main is 100.6 % (PASS via G3 band). Branch "
+    "is strictly better than main yet rejected by the strict 90 % gate. "
+    "Pick the smallest contract reconciliation to unblock CI."
+)
+
+REVIEW_PROMPT = """\
+# Council Adjudication — Post-Revert Always-Budget Reconciliation
+
+## Current state (after R4 revert is committed locally)
+
+```
+FAIL  always-rule extended budget: 47,448 / 49,000 chars (96.8%)
+      thresholds: warn 80% · fail 90% · per-rule ≤ 6,000 (ext) ·
+                  top-3 ≤ 24,500 (ext) · depth ≤ 2 · G3 band ≤ +2%
+```
+
+`main` (last green): **49,311 / 49,000 = 100.6 %** — passes because
+the linter only rejects 90–100 % (gap zone) and > 102 %; 100–102 %
+is the documented G3 tolerance band.
+
+This branch is objectively closer to the cap than `main` but
+rejected. The 90 % gate was written under the assumption Phase 2A
+would land below 90 %. Phase 2A is structurally infeasible under
+model (b) (prior council finding); roadmap re-routes the trim to
+Phase 5 via either model (c) shared-divisor or shared-mechanics
+consolidation.
+
+## Fixed option set (pick exactly one)
+
+- **A1 — Recovery-band carve-out.** Extend the linter so "current
+  pct ≤ last-green-`main` pct AND pct ≤ 100 %" becomes a documented
+  `WARN (recovery band)`, not FAIL. Persists until Phase 5
+  re-tightens. No baseline regression possible; needs `main`-pct
+  lookup (committed scalar in repo or CI artifact).
+- **A2 — Lower FAIL_THRESHOLD to 100 %.** Drop the 90 % gate to
+  match the contract's spirit ("< cap is acceptable; > cap +2 %
+  fails"). Phase 5 re-tightens to 85 %/95 % per its own task list.
+  Smallest diff. Risk: removes the early-warning gate.
+- **A3 — Trim ~3.4k chars from current rules.** Find another
+  candidate (none of the three Phase-2A targets, since 2A proved
+  extraction adds overhead). Risk: the next-largest rules touch
+  the safety floor (Q3=A locked).
+- **A4 — Bump TOTAL_CAP to 53,000 + keep 90 % FAIL.** Rewrites the
+  contract's total cap upward; freezes current state at ~89.5 %.
+  Defers Phase 5 trim to a separate roadmap. Cleanest CI but
+  contract drift.
+- **A5 — Mark phase as `irreducible-at-current-method` and freeze
+  the linter at current state.** Pin a `CURRENT_BASELINE = 47_448`
+  ceiling alongside the 90 % gate; PASS while pct ≤ baseline,
+  FAIL on regression. Deflects Phase 5 work upstream into model (c).
+
+## Constraints
+
+- Safety-floor rules (`non-destructive-by-default`, `commit-policy`,
+  `scope-control`, `verify-before-complete`) are Q3=A locked and
+  may not be slimmed.
+- `road-to-structural-optimization` Phase 5 owns the long-term
+  trim — do not duplicate work into Phase 2A.
+- The four other phases (1, 2B, 3, 4, 6) on this branch are
+  blocked by this CI gate; the choice gates roadmap throughput.
+
+## Output Contract (STRICT)
+
+```
+### Verdict
+**Choice:** <A1 | A2 | A3 | A4 | A5>
+**One-sentence rationale:** <≤ 30 words>
+```
+
+```
+### Required follow-up actions (numbered, ≤ 4)
+  1. <smallest concrete step, files named>
+  2. <...>
+```
+
+```
+### Risk note
+**Single biggest risk of the chosen option:** <one sentence>
+**Mitigation:** <one sentence>
+```
+
+```
+### Contract amendment needed?
+**Amend load-context-budget-model.md?** <YES — section · NO>
+**Amend road-to-structural-optimization Phase 5 success criterion?**
+  <YES — new criterion in 1 line · NO>
+```
+
+Be decisive — total response ≤ 800 words. Artefacts follow verbatim.
+"""
+
+
+def main() -> int:
+    anthropic = AnthropicClient(api_key=load_anthropic_key(), model="claude-sonnet-4-5")
+    openai = OpenAIClient(api_key=load_openai_key(), model="gpt-4o")
+    members = [anthropic, openai]
+
+    context = bundle_files(ARTEFACTS)
+    project = detect_project_context(REPO_ROOT)
+    table = load_prices()
+
+    user_prompt = REVIEW_PROMPT + "\n\n---\n\n" + context.text
+
+    question = CouncilQuestion(mode="files", user_prompt=user_prompt, max_tokens=2048)
+    estimates = estimate(question, members, table, project=project, original_ask=ORIGINAL_ASK)
+    print("=== ESTIMATE (single round) ===")
+    total_est = 0.0
+    for c, e in zip(members, estimates):
+        print(f"  {c.name}/{c.model}: ~{e.input_tokens} in + {e.output_tokens} out = ${e.total_usd:.4f}")
+        total_est += e.total_usd
+    print(f"  TOTAL per round (max): ${total_est:.4f}")
+
+    budget = CostBudget(max_input_tokens=200_000, max_output_tokens=80_000, max_calls=20, max_total_usd=2.50)
+    rounds_collected: list[list] = []
+
+    def _on_round_complete(round_idx, round_responses) -> None:
+        rounds_collected.append(list(round_responses))
+        print(f"=== ROUND {round_idx + 1} COMPLETE ===")
+        for r in round_responses:
+            if r.error:
+                print(f"  [error] {r.provider}/{r.model}: {r.error}")
+                continue
+            a = estimate_cost(r.provider, r.model, r.input_tokens, r.output_tokens, table)
+            print(f"  [done] {r.provider}/{r.model}: {r.input_tokens} in / {r.output_tokens} out · ${a.total_usd:.4f}")
+
+    print("=== CONSULT (1 round, post-revert reconciliation) ===")
+    consult(members, question, budget, rounds=1, on_round_complete=_on_round_complete,
+            table=table, project=project, original_ask=ORIGINAL_ASK)
+    if not rounds_collected:
+        return 1
+    actual = sum(estimate_cost(r.provider, r.model, r.input_tokens, r.output_tokens, table).total_usd
+                 for rr in rounds_collected for r in rr if not r.error)
+    print(f"=== TOTAL ACTUAL: ${actual:.4f} ===")
+    final = rounds_collected[-1]
+    if not [r for r in final if not r.error]:
+        return 1
+    manifest = SessionManifest(
+        mode="files", artefact="agents/roadmaps/road-to-structural-optimization.md",
+        original_ask=ORIGINAL_ASK, members=[f"{r.provider}/{r.model}" for r in final],
+        rounds=len(rounds_collected), cost_usd_estimated=total_est, cost_usd_actual=actual,
+        extra={"purpose": "Post-revert always-budget contract reconciliation"},
+    )
+    sd = save_session(manifest=manifest, responses=rounds_collected)
+    print(f"[saved] {sd.relative_to(REPO_ROOT)}/")
+    return 0
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())

package/scripts/ai_council/_one_off_rebalancing_audit.py

@@ -0,0 +1,149 @@
+"""One-off council consultation — Phase 0 audit findings on rebalancing roadmap.
+
+Validates whether the rebalancing roadmap premise still holds against the
+actual PR #34 diff. Transient script; can be deleted after the consult runs.
+
+Invocation:
+    .venv/bin/python -m scripts.ai_council._one_off_rebalancing_audit
+"""
+from __future__ import annotations
+
+import sys
+from pathlib import Path
+
+from scripts.ai_council.clients import AnthropicClient, load_anthropic_key
+from scripts.ai_council.orchestrator import CostBudget, CouncilQuestion, consult, estimate
+from scripts.ai_council.pricing import estimate_cost, load_prices
+from scripts.ai_council.project_context import detect_project_context
+
+REPO_ROOT = Path(__file__).resolve().parents[2]
+
+ORIGINAL_ASK = (
+    "Phase 0 audit of the rebalancing roadmap. The roadmap was written based on "
+    "five rounds of external review claiming PR #34 over-deleted implicit "
+    "expertise. Validate whether the premise still holds against the actual diff. "
+    "Three concrete questions: (1) is the deletion narrative supported by the "
+    "numbers? (2) which of the 6 phases are already done or moot? "
+    "(3) what is the actual minimum-viable scope still worth executing?"
+)
+
+ARTEFACT = """# Phase 0 audit findings - road-to-rebalancing.md
+
+## Premise from the roadmap
+
+The risk surface is whether implicit expertise (edge cases, decision forks,
+failure modes, anti-patterns) was trimmed alongside the redundancy.
+Rebalancing means restoring intelligence without re-inflating Always-rules.
+
+## Actual numbers from the PR #34 diff
+
+Scope: git diff origin/main...HEAD, path .agent-src.uncompressed/rules/
+
+- 35 files changed: 202 insertions, 204 deletions => net -2 lines total.
+- Largest delta: language-and-tone.md 37 ins / 96 del. The 96 lines were
+  EXTRACTED to docs/guidelines/language-and-tone-examples.md (79 lines),
+  not deleted. Net knowledge loss: ~17 lines of duplicated phrasing.
+- Second-largest: roadmap-progress-sync.md 26 ins / 33 del - minor.
+- All other rules: <=8 line changes each, mostly renames.
+- ZERO rule files deleted outright.
+
+## Phase-by-phase realities
+
+### Phase 0 - Removed-Knowledge Audit
+This document IS Phase 0. Findings: 80% redundancy, examples extracted to
+safe layer, zero decision-logic deletions. Phase 0 is now COMPLETE.
+
+### Phase 1 - Pilot Context Split (3 rules)
+Candidates: autonomous-execution, minimal-safe-diff, scope-control.
+- autonomous-execution: 8 line delta in PR, no extraction yet.
+- minimal-safe-diff: already auto-trigger; 6 line delta.
+- scope-control: 40 line ADDITION (not deletion) in this PR. Pilot value low.
+
+### Phase 2 - load_context: convention + linter
+0 rules use load_context: today. Convention does not exist.
+Genuinely net-new work.
+
+### Phase 3 - Guidelines domain folders
+Already done. 47 guidelines, 46 already in domain folders
+(agent-infra/, docs/, e2e/, php/). Only language-and-tone-examples.md
+is flat at root. Phase 3 reduces to deciding where the one flat file goes.
+
+### Phase 4 - Golden-Transcript-backed demos under examples/flows/
+Partially shipped. docs/end-to-end-walkthroughs.md (built last cycle)
+already does this with 4 traces anchored to GT-1, GT-P1, GT-U2, GT-2.
+Required cases per roadmap: implement-ticket-demo, work-freeform-demo,
+ui-track-demo, blocked-path-demo (all covered) plus mixed-flow-demo
+(NOT covered yet).
+Net-new: 1 mixed-flow demo + folder move from docs/ to examples/flows/.
+
+### Phase 5 - Rule priority hierarchy + interaction matrix
+Partially shipped. docs/contracts/rule-interactions.yml and
+rule-interactions.md exist (13 pairs across 9 rules).
+rule-priority-hierarchy.md does NOT exist.
+
+### Phase 6 - Senior-agent behavior tests
+Not started. Net-new work, but only valuable if Phase 1 actually runs
+and produces something to validate.
+
+## Question to council
+
+Given:
+- The deletion narrative is empirically thin (-2 net lines after extraction).
+- 3 of 6 phases are already done or near-done.
+- Phase 1 pilot value is questionable on the 3 named candidates.
+
+Should this roadmap:
+
+(A) Close out as substantially-already-done. Execute only the small delta
+    (1 mixed-flow demo, move walkthroughs to examples/flows/, add the
+    priority-hierarchy doc, add load_context: convention if cheap),
+    then archive.
+
+(B) Drop Phase 1 pilot - the three named candidates don't show evidence of
+    over-deletion. Execute Phase 2 + Phase 4 (1 missing demo + restructure)
+    + Phase 5 (priority-hierarchy doc) only.
+
+(C) Original full scope - assume the audit missed something subtle and run
+    all 6 phases as written.
+
+(D) Other framing - propose a tighter scope based on what the audit shows.
+
+Identify blind spots: did the diff miss content moves between branches?
+Are there rules whose implicit expertise lives in non-line-count signal?
+Recommend (A/B/C/D) with rationale.
+"""
+
+
+def main() -> int:
+    api_key = load_anthropic_key()
+    client = AnthropicClient(api_key=api_key)
+    project = detect_project_context(REPO_ROOT)
+    table = load_prices()
+
+    question = CouncilQuestion(mode="roadmap", user_prompt=ARTEFACT, max_tokens=2048)
+    estimates = estimate(question, [client], table, project=project, original_ask=ORIGINAL_ASK)
+    print(f"[estimate] ~{estimates[0].input_tokens} in + {estimates[0].output_tokens} out = ${estimates[0].total_usd:.4f}")
+
+    budget = CostBudget(
+        max_input_tokens=50_000, max_output_tokens=20_000,
+        max_calls=5, max_total_usd=0.50,
+    )
+
+    print(f"[consult] calling {client.name}/{client.model} ...")
+    responses = consult([client], question, budget, table=table, project=project, original_ask=ORIGINAL_ASK)
+    if not responses or responses[0].error:
+        err = responses[0].error if responses else "no response"
+        print(f"[error] {err}", file=sys.stderr)
+        return 1
+
+    r = responses[0]
+    actual = estimate_cost(r.provider, r.model, r.input_tokens, r.output_tokens, table)
+    print(f"[done] {r.input_tokens} in / {r.output_tokens} out, {r.latency_ms} ms, actual ${actual.total_usd:.4f}")
+    print("=" * 72)
+    print(r.text)
+    print("=" * 72)
+    return 0
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())