@event4u/agent-config 6.0.0 → 7.0.0

package/src/scripts/ai_council/one_off_archive/2026-05/_one_off_context_layer_v1_review.py

@@ -1,292 +0,0 @@
-"""Council review of road-to-context-layer-maturity.md draft v1 + PR #36.
-
-Both reviewers ran independent preview-reviews on PR #36 and converged on
-the same gap: rule layer is mature, context layer is unproven. The host
-agent distilled the convergence into road-to-context-layer-maturity.md
-(draft v1, 6 phases, lightweight tier).
-
-Council task: validate v1 against (a) the two preview-reviews summarised
-inline, (b) the actual PR #36 shape (description + diff stat), and (c)
-the existing always-budget contract. Together with the host agent,
-define the binding step list before lock.
-
-Single round, structured per-phase verdict. Saves the session under
-agents/council-sessions/.
-
-Invocation:
-    .venv/bin/python -m scripts.ai_council._one_off_context_layer_v1_review
-"""
-from __future__ import annotations
-
-import subprocess
-import sys
-from pathlib import Path
-
-from scripts.ai_council.bundler import bundle_prompt
-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[3]
-ROADMAP_PATH = REPO_ROOT / "agents/roadmaps/road-to-context-layer-maturity.md"
-
-ORIGINAL_ASK = (
-    "Two independent preview-reviews on PR #36 converged on the same "
-    "finding: rule layer is now mature; context layer is unproven. The "
-    "host agent distilled the convergence into a 6-phase lightweight "
-    "roadmap (road-to-context-layer-maturity.md draft v1). Council task: "
-    "validate v1 against the reviewers' findings, the PR's actual shape, "
-    "and the always-budget contract; together with the host agent, "
-    "define the binding step list."
-)
-
-REVIEW_PROMPT_HEADER = """\
-# Council Review — road-to-context-layer-maturity.md draft v1 + PR #36
-
-## Background (verbatim, do not re-frame)
-
-Two preview-reviews of PR #36 (`feat/better-basement` branch, 174 files,
-+20k/-3.6k diff, structural-optimization roadmap closed + 1.16.0
-follow-ups Phase 1+2 closed) produced converging findings.
-
-### Reviewer #1 — 3-Layer Architecture lens
-
-- PR #36 is not a feature PR; it is the first **post-maturity
-  architecture move**.
-- Names a 3-Layer Architecture: **Rule** (Obligation / MUST), **Context**
-  (Decision Logic / HOW to think), **Examples** (Pattern Memory / HOW it
-  looks). Phase 2A revert proved the boundary matters — moving Iron-Law
-  prose into context made the rule unsafe.
-- Lauds: contexts/communication/ as a previously-unsolved area;
-  always-budget guard + golden tests as a maturity signal.
-- Gaps: (a) no Level-2 context-loading design (chain, priority,
-  budget, activation order); (b) no outcome measurement — golden tests
-  measure structure, not whether the agent decides better; (c)
-  context-activation clarity unclear (when, how many concurrently);
-  (d) Iron Laws must stay in Rule, never migrate to Context; (e)
-  Examples-as-demos under-leveraged.
-- Verdict: 9.6/10. Recommendation: do NOT add features/skills/rules;
-  perfect Context, make Examples real demos, make Decision-System
-  visible.
-
-### Reviewer #2 — Consolidation / scope-discipline lens
-
-- PR #36 is correctly framed as "Structural optimization foundation +
-  regression gates + command surface reduction", not a feature PR.
-- Lauds: Phase 2A abort + honest Model-(b) finding (Context Tax >
-  rule-slim gain); roadmap reflexion depth (council rounds, locked
-  decisions, file-ownership matrix); test gates; 1.16-followup clean
-  closure.
-- Gaps: (1) PR size 174 files reduces reviewability; (2) roadmaps
-  too heavy — auditable but not consumable for normal feature work,
-  needs complexity standard; (3) always-budget headroom 1,552 chars
-  (96.8% utilization) — slimming hebel exhausted, new strategy needed
-  (demote / merge / hard-condense / shared-context amortization);
-  (4) slow-rollout protocol condensed under autonomy mandate;
-  (5) one-off scripts accumulating in scripts/ai_council/.
-- P0 before merge: PR description rewrite, one-off-script decision,
-  honest budget block, Phase 2A as finding.
-- Verdict: 8.8/10 current, 9.1/10 with cleanup. "Inhaltlich stark.
-  Strategisch richtig. Aber zu groß und zu meta-lastig."
-
-### Convergence
-
-- Both: don't add more rules / skills / commands.
-- Both: context activation / loading / budget = the unsolved next
-  question.
-- Both: praise audit honesty, criticise size/concentration.
-- Both: PR #36 should ship after cleanup, not be expanded.
-
-## PR #36 shape
-
-- 174 files changed · +20,120 / −3,629
-- Branch: feat/better-basement → main (open)
-- Title: docs(roadmap): add structural-optimization v3.1 + 1.16.0 follow-ups v1.1
-
-(Full diff stat appended below; PR body appended below.)
-
-## Your task
-
-Review **road-to-context-layer-maturity.md draft v1** (full text appended
-below) against:
-
-1. **Reviewer convergence:** does v1 close every gap both reviewers
-   named, or does it merely paraphrase them?
-2. **PR #36 fit:** is Phase 0 (PR closeout) sufficient to ship #36
-   honestly, or are items missing that the diff exposes?
-3. **Always-budget contract:** is Phase 4 (Always-Budget v2)
-   structurally sound, or does it re-walk Phase 2A's path?
-4. **Lightweight-tier discipline:** does the roadmap itself follow
-   the standard it locks in Phase 5, or does it cheat its own gate?
-5. **Sequencing:** is `0 → 1 → 2 → 3 → 4 → 5` the right order, or
-   should something move?
-
-## Output contract (STRICT)
-
-For EACH of the six phases, produce:
-
-```
-### Phase N — <title>
-
-**Verdict:** <ACCEPT | PARTIAL | REJECT>
-**What v1 gets right (1 sentence):** ...
-**What v1 misses or over-reaches (1-2 sentences):** ...
-**Concrete change to v2 (binding):** ...
-```
-
-Then a final block:
-
-```
-### Greenlight verdict
-
-<one of: FULL GREENLIGHT — lock v1 / CONDITIONAL GREENLIGHT — apply N
-revisions then lock / BLOCKED — major rework needed>
-
-**Binding revisions for v2 (numbered, ≤ 8):** ...
-**Estimated total effort:** <engineer-days>
-**One-line strategic risk you would still fly with:** ...
-```
-
-Total response budget: ≤ 1800 words. Do not re-write the roadmap. Do
-not propose a separate roadmap unless you mark BLOCKED.
-"""
-
-
-def _read(path: Path) -> str:
-    return path.read_text(encoding="utf-8") if path.exists() else ""
-
-
-def _diff_stat() -> str:
-    proc = subprocess.run(
-        ["git", "diff", "--stat", "origin/main..HEAD"],
-        cwd=REPO_ROOT, check=True, capture_output=True, text=True,
-    )
-    lines = proc.stdout.splitlines()
-    if len(lines) > 60:
-        return "\n".join(lines[:30] + ["... [middle truncated] ..."] + lines[-15:])
-    return proc.stdout
-
-
-def _pr_body() -> str:
-    try:
-        proc = subprocess.run(
-            ["gh", "pr", "view", "36", "--json", "body", "--jq", ".body"],
-            cwd=REPO_ROOT, check=True, capture_output=True, text=True,
-        )
-        return proc.stdout
-    except subprocess.CalledProcessError:
-        return "[gh pr view 36 unavailable]"
-
-
-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]
-
-    roadmap_text = _read(ROADMAP_PATH)
-    if not roadmap_text:
-        print(f"[error] roadmap not found: {ROADMAP_PATH}", file=sys.stderr)
-        return 1
-
-    bundle_text = "\n\n---\n\n".join([
-        REVIEW_PROMPT_HEADER,
-        "## PR #36 — diff --stat (origin/main..HEAD)\n\n```\n" + _diff_stat() + "\n```",
-        "## PR #36 — body (verbatim)\n\n" + _pr_body(),
-        "## Roadmap v1 (verbatim, the artefact to validate)\n\n" + roadmap_text,
-    ])
-    context = bundle_prompt(bundle_text)
-    project = detect_project_context(REPO_ROOT)
-    table = load_prices()
-
-    question = CouncilQuestion(
-        mode="prompt",
-        user_prompt=context.text,
-        max_tokens=4096,
-    )
-
-    estimates = estimate(
-        question, members, table, project=project, original_ask=ORIGINAL_ASK,
-    )
-    print("=== ESTIMATE (single round, max tokens) ===")
-    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) ===")
-    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="prompt",
-        artefact=str(ROADMAP_PATH.relative_to(REPO_ROOT)),
-        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": "Council review of road-to-context-layer-maturity v1 + PR #36"},
-    )
-    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/src/scripts/ai_council/one_off_archive/2026-05/_one_off_followups_review.py

@@ -1,259 +0,0 @@
-"""Council review of road-to-1-16-followups.md (draft v1).
-
-The followups roadmap distils 14 numbered findings (F1-F14) from two
-external review rounds of release 1.16.0 into three phases:
-- Phase 0: reviewer P0/P1/P2 stack (README sync, budget headroom,
-  council labelling, release-tag gate). Effort <= 1 hour.
-- Phase 1: load_context: rollout to three medium-risk policy rules
-  + non-destructive-by-default failure-mode-only split + 1.15-followups
-  archive verification. Effort ~2 days.
-- Phase 2: README "Start here" anchor, host-agent wording precision,
-  golden-test failure-mode coverage. Effort ~0.5 days.
-
-Single round. Both members produce a structured per-phase verdict and
-together with the host agent define the binding step list before the
-roadmap is locked.
-
-Invocation:
-    .venv/bin/python -m scripts.ai_council._one_off_followups_review
-"""
-from __future__ import annotations
-
-import sys
-from pathlib import Path
-
-from scripts.ai_council.bundler import bundle_roadmap
-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[3]
-ROADMAP_PATH = REPO_ROOT / "agents/roadmaps/road-to-1-16-followups.md"
-
-ORIGINAL_ASK = (
-    "Two external review rounds on release 1.16.0 produced 14 numbered "
-    "findings (F1-F14) and a P0/P1/P2 fix stack. The host agent distilled "
-    "them into road-to-1-16-followups.md (draft v1, 3 phases, 7 risks). "
-    "Council task: evaluate the F1-F14 capture quality, the phase split, "
-    "the gating between this roadmap and road-to-structural-optimization "
-    "v3.1, and the per-step contracts. Together with the host agent, "
-    "define the binding step list before lock-in."
-)
-
-REVIEW_PROMPT = """\
-# Council Review — road-to-1-16-followups.md draft v1
-
-Context for both reviewers:
-
-- A separate roadmap (`road-to-structural-optimization.md`) was \
-  finalised at v3.1 in a prior session. You both reviewed it across \
-  five rounds. v3.1 is locked. **Do not re-litigate it.** This review \
-  is about a *different*, *narrower* roadmap that addresses post-1.16.0 \
-  PR feedback.
-- Reviewer 1 (consolidation-quality lens) scored 1.16.0 at 9.5/10. \
-  Concerns: surface still large, README dense, context-layer rollout \
-  pace, "agent runs / implements" overclaim risk, golden-test failure-\
-  mode bias. Strategy notes (F5-F8): roll out load_context: slowly, \
-  prepare a second command-collapse phase, ship outcome demos, do not \
-  inflate the README.
-- Reviewer 2 (commit-level audit, raw.githubusercontent.com verified) \
-  scored 1.16.0 at A-. Concrete findings:
-  - **F9 (P0):** README on `main` shows pre-1.15.0 wording \
-    ("Teach your AI agents Laravel...", "124 Skills · 46 Rules · \
-    73 Commands · 46 Guidelines"). The 1.16.0 release tag has the \
-    correct text but is not an ancestor of `main`. Reviewer-supplied \
-    cherry-pick SHAs: 1053d56, d26bf68, 2fa8022, c282ae3.
-  - **F10 (P1):** `tests/test_always_budget.py` top-5 cap headroom is \
-    ~141 chars after `no-cheap-questions` was added. Target: >= 2,000 \
-    chars headroom. Q3=A from structural-optimization v3.1 excludes \
-    safety-floor rules from slim work.
-  - **F11 (P2):** AI Council Phase 3-4 (debate, persistence, special \
-    modes) grew without external-user signal. Reviewer fix: experimental \
-    banner, no code cut.
-  - **F14:** 1.16.0 tag points to commit 08daac9 which is not on `main`. \
-    Workflow gap.
-- Three Reviewer-1 strategy items (F5/F6/F7) and one (F8) on golden \
-  tests are also captured. The host agent placed them as: F5/F6/F7 = \
-  out-of-scope strategy notes (next-roadmap-generation seed); F8 = \
-  Phase 2.3.
-
-You both reviewed v3.1 already. **Stop re-reviewing v3.1.** Pick a \
-binding verdict per phase of *this* followups roadmap, and together \
-with the host agent define the final step list.
-
-## Output Contract (STRICT)
-
-For each phase, produce exactly this block:
-
-```
-### Phase <0|1|2>: <theme>
-
-**Verdict:** <ACCEPT | ACCEPT_WITH_REVISIONS | REJECT>
-**Capture quality (F-items folded into this phase):** <COMPLETE | PARTIAL — list missing> 
-**Gate correctness:** <CORRECT | UNDER-GATED | OVER-GATED>
-**Per-step contract clarity:** <SUFFICIENT | NEEDS_TIGHTENING — name steps>
-**Required revisions (numbered, 1-3 max):**
-  1. <one sentence — smallest change>
-  2. <...>
-  3. <...>
-**Risk register coverage:** <SUFFICIENT | MISSING — name risk>
-```
-
-Verdict definitions:
-- **ACCEPT** — phase ships as written; no revisions required.
-- **ACCEPT_WITH_REVISIONS** — phase ships after the listed revisions \
-  land. Each revision must be one specific, actionable change.
-- **REJECT** — phase is structurally wrong; describe the structural \
-  fault in <= 2 sentences in the revisions block.
-
-## The Three Phases
-
-The roadmap text follows this prompt verbatim. Read it first, then \
-render the three blocks above for Phase 0, Phase 1, Phase 2.
-
-## Cross-cutting questions (answer after the three blocks)
-
-```
-### Cross-cutting
-
-**Is the F1-F14 placement correct?** \
-  <YES | NO — list misplaced F-numbers and where they should go>
-**Is the gating to road-to-structural-optimization v3.1 correct?** \
-  <YES | NO — Phase 1 currently gates on v3.1 Phase 0.4 worked example>
-**Is anything missing entirely from the F1-F14 capture that the \
-  source feedback raised?** \
-  <NO | YES — list missing observation + which phase/risk it belongs to>
-**Is anything over-scoped (should not be a roadmap item at all)?** \
-  <NO | YES — list and justify>
-**Reviewer-1 strategy items F5/F6/F7 placement (out-of-scope, \
-  next-roadmap-generation seed) — correct?** \
-  <YES | NO — propose alternative placement>
-```
-
-## Final Output
-
-After the three phase blocks and the cross-cutting block, add:
-
-```
-### Consensus verdict
-
-**Overall recommendation:** <one of: LOCK_AS_IS | LOCK_AFTER_REVISIONS \
-  | RESCOPE_REQUIRED>
-**Phase-0 launchable now?** <YES | NO — reasons>
-**Phase-1 unblocked once structural v3.1 Phase 0.4 lands?** \
-  <YES | NO — reasons>
-**Total residual revisions:** <count and sum of effort estimates>
-**Single biggest risk in this followups roadmap that the host agent \
-  has under-weighted:** <one sentence>
-```
-
-Be decisive. The host agent will integrate your two verdicts; \
-divergence triggers a tie-break round, but only if you actually \
-disagree on a *blocking* item. Total response budget: <= 1500 words \
-per reviewer.
-"""
-
-
-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_roadmap(ROADMAP_PATH)
-    project = detect_project_context(REPO_ROOT)
-    table = load_prices()
-
-    user_prompt = REVIEW_PROMPT + "\n\n---\n\n" + context.text
-
-    question = CouncilQuestion(
-        mode="roadmap",
-        user_prompt=user_prompt,
-        max_tokens=4096,
-    )
-
-    estimates = estimate(
-        question, members, table, project=project, original_ask=ORIGINAL_ASK,
-    )
-    print("=== ESTIMATE (single round, max tokens) ===")
-    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, followups roadmap review) ===")
-    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="roadmap",
-        artefact=str(ROADMAP_PATH.relative_to(REPO_ROOT)),
-        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": "Council review of road-to-1-16-followups.md draft v1"},
-    )
-    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/src/scripts/ai_council/one_off_archive/2026-05/_one_off_inject_quiet_flag.py

@@ -1,33 +0,0 @@
-#!/usr/bin/env python3
-"""One-off: inject {{.QUIET_FLAG}} into Taskfile cmds for --quiet-aware scripts."""
-from __future__ import annotations
-import re
-from pathlib import Path
-
-# Scripts that accept --quiet (verified: 20 total = 3 pre-existing + 17 patched).
-QUIET_AWARE = {
-    "check_always_budget", "check_one_off_location", "check_safety_floor_untouched",
-    "check_augmentignore", "check_command_count_messaging", "check_condensed_paths",
-    "check_council_layout", "check_council_references", "check_iron_law_prominence",
-    "check_md_language", "check_memory_proposal", "check_public_catalog_links",
-    "check_reply_consistency", "check_roadmap_trackable",
-    "lint_examples", "lint_handoffs", "lint_load_context", "lint_roadmap_complexity",
-    "lint_rule_interactions", "lint_rule_tiers",
-}
-
-TASKFILES = sorted(Path("taskfiles").glob("*.yml")) + [Path("Taskfile.yml")]
-patched = 0
-for tf in TASKFILES:
-    text = tf.read_text()
-    new = text
-    for name in QUIET_AWARE:
-        # Match: cmd: python3 scripts/<name>.py [args]
-        # Insert {{.QUIET_FLAG}} after the script path, before any other args.
-        pat = re.compile(rf"(python3 scripts/{name}\.py)(\s|$)(?!.*QUIET_FLAG)")
-        new, n = pat.subn(r"\1 {{.QUIET_FLAG}}\2", new)
-        if n:
-            patched += n
-    if new != text:
-        tf.write_text(new)
-        print(f"  patched: {tf}")
-print(f"\nTotal injections: {patched}")

package/src/scripts/ai_council/one_off_archive/2026-05/_one_off_measure_v2.sh

@@ -1,36 +0,0 @@
-#!/bin/bash
-# Phase 10.7 baseline — runs only verbosity-aware patched-script tasks.
-# Skips broken-on-dirty-tree tasks (consistency, check-index, validate-schema).
-# This subset is what actually demonstrates the --quiet effect.
-TASKS="check-condensed-paths check-refs check-portability lint-roadmap-complexity check-public-catalog-links check-command-count check-cluster-patterns lint-rule-interactions lint-load-context check-context-paths check-no-roadmap-refs check-council-references lint-one-off-age check-reply-consistency check-iron-law-prominence check-always-budget check-one-off-location lint-rule-budget lint-skills lint-rule-tiers lint-handoffs lint-marketplace lint-examples"
-
-run() {
-  local label=$1
-  local level=$2
-  local out=$3
-  AGENT_SCRIPT_VERBOSITY=$level task $TASKS > "$out" 2>&1
-  local lines=$(wc -l < "$out" | tr -d ' ')
-  local chars=$(wc -c < "$out" | tr -d ' ')
-  echo "$label: lines=$lines chars=$chars"
-}
-
-echo "=== MINIMAL ==="
-run "MINIMAL" minimal /tmp/ci-min.log
-echo ""
-echo "=== VERBOSE ==="
-run "VERBOSE" verbose /tmp/ci-vrb.log
-
-ML=$(wc -l < /tmp/ci-min.log | tr -d ' ')
-MC=$(wc -c < /tmp/ci-min.log | tr -d ' ')
-VL=$(wc -l < /tmp/ci-vrb.log | tr -d ' ')
-VC=$(wc -c < /tmp/ci-vrb.log | tr -d ' ')
-
-echo ""
-python3 -c "
-ml=$ML; vl=$VL; mc=$MC; vc=$VC
-dl=(vl-ml)/vl*100 if vl else 0
-dc=(vc-mc)/vc*100 if vc else 0
-print(f'Lines: {ml} -> {vl}, reduction={dl:.1f}% (target >=40%)')
-print(f'Chars: {mc} -> {vc}, reduction={dc:.1f}%')
-print('verdict:', 'MET' if dl >= 40 else f'MISSED ({dl:.1f}% < 40%)')
-"

package/src/scripts/ai_council/one_off_archive/2026-05/_one_off_measure_verbosity.sh

@@ -1,26 +0,0 @@
-#!/bin/bash
-# Phase 10.7 baseline measurement: post-consistency task ci subset.
-TASKS="counts-check check-index check-router check-condensation check-condensed-paths check-refs check-token-optimizer-freshness check-portability check-examples-shape lint-roadmap-complexity check-public-links check-public-catalog-links check-command-count lint-no-new-atomic-commands check-cluster-patterns lint-rule-interactions lint-load-context check-context-paths check-no-roadmap-refs check-council-references lint-one-off-age check-ownership-matrix check-reply-consistency check-iron-law-prominence check-always-budget check-one-off-location validate-schema lint-rule-budget lint-skills lint-rule-tiers lint-handoffs lint-hook-manifest lint-showcase-sessions lint-marketplace"
-
-echo "=== MINIMAL ==="
-AGENT_SCRIPT_VERBOSITY=minimal task $TASKS 2>&1 > /tmp/ci-min.log
-ML=$(wc -l < /tmp/ci-min.log | tr -d ' ')
-MC=$(wc -c < /tmp/ci-min.log | tr -d ' ')
-echo "MINIMAL: lines=$ML chars=$MC"
-
-echo ""
-echo "=== VERBOSE ==="
-AGENT_SCRIPT_VERBOSITY=verbose task $TASKS 2>&1 > /tmp/ci-vrb.log
-VL=$(wc -l < /tmp/ci-vrb.log | tr -d ' ')
-VC=$(wc -c < /tmp/ci-vrb.log | tr -d ' ')
-echo "VERBOSE: lines=$VL chars=$VC"
-
-echo ""
-python3 -c "
-ml=$ML; vl=$VL; mc=$MC; vc=$VC
-dl=(vl-ml)/vl*100 if vl else 0
-dc=(vc-mc)/vc*100 if vc else 0
-print(f'Lines: {ml} -> {vl}, reduction={dl:.1f}% (target >=40%)')
-print(f'Chars: {mc} -> {vc}, reduction={dc:.1f}%')
-print('verdict:', 'MET' if dl >= 40 else f'MISSED ({dl:.1f}% < 40%)')
-"