devlyn-cli 2.2.2 → 2.3.1
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/AGENTS.md +2 -2
- package/CLAUDE.md +4 -4
- package/README.md +85 -34
- package/benchmark/auto-resolve/BENCHMARK-DESIGN.md +61 -44
- package/benchmark/auto-resolve/BENCHMARK-RESULTS.md +341 -0
- package/benchmark/auto-resolve/README.md +307 -44
- package/benchmark/auto-resolve/RUBRIC.md +23 -14
- package/benchmark/auto-resolve/fixtures/F1-cli-trivial-flag/NOTES.md +7 -3
- package/benchmark/auto-resolve/fixtures/F10-persist-write-collision/NOTES.md +8 -3
- package/benchmark/auto-resolve/fixtures/F11-batch-import-all-or-nothing/NOTES.md +8 -3
- package/benchmark/auto-resolve/fixtures/F12-webhook-raw-body-signature/NOTES.md +10 -4
- package/benchmark/auto-resolve/fixtures/F15-frozen-diff-race-review/NOTES.md +10 -4
- package/benchmark/auto-resolve/fixtures/F16-cli-quote-tax-rules/NOTES.md +12 -0
- package/benchmark/auto-resolve/fixtures/F16-cli-quote-tax-rules/spec.md +6 -0
- package/benchmark/auto-resolve/fixtures/F2-cli-medium-subcommand/NOTES.md +7 -4
- package/benchmark/auto-resolve/fixtures/F21-cli-scheduler-priority/NOTES.md +12 -0
- package/benchmark/auto-resolve/fixtures/F21-cli-scheduler-priority/spec.md +6 -0
- package/benchmark/auto-resolve/fixtures/F22-cli-ledger-close/NOTES.md +8 -0
- package/benchmark/auto-resolve/fixtures/F23-cli-fulfillment-wave/NOTES.md +12 -0
- package/benchmark/auto-resolve/fixtures/F23-cli-fulfillment-wave/spec.md +6 -0
- package/benchmark/auto-resolve/fixtures/F25-cli-cart-promotion-rules/NOTES.md +16 -4
- package/benchmark/auto-resolve/fixtures/F25-cli-cart-promotion-rules/spec.md +7 -0
- package/benchmark/auto-resolve/fixtures/F26-cli-payout-ledger-rules/NOTES.md +11 -5
- package/benchmark/auto-resolve/fixtures/F3-backend-contract-risk/NOTES.md +8 -1
- package/benchmark/auto-resolve/fixtures/F3-backend-contract-risk/expected.json +4 -2
- package/benchmark/auto-resolve/fixtures/F3-backend-contract-risk/spec.md +1 -1
- package/benchmark/auto-resolve/fixtures/F31-cli-seat-rebalance/NOTES.md +34 -0
- package/benchmark/auto-resolve/fixtures/F31-cli-seat-rebalance/expected.json +57 -0
- package/benchmark/auto-resolve/fixtures/F31-cli-seat-rebalance/metadata.json +10 -0
- package/benchmark/auto-resolve/fixtures/F31-cli-seat-rebalance/setup.sh +2 -0
- package/benchmark/auto-resolve/fixtures/F31-cli-seat-rebalance/spec.md +67 -0
- package/benchmark/auto-resolve/fixtures/F31-cli-seat-rebalance/task.txt +7 -0
- package/benchmark/auto-resolve/fixtures/F31-cli-seat-rebalance/verifiers/duplicate-event-error.js +35 -0
- package/benchmark/auto-resolve/fixtures/F31-cli-seat-rebalance/verifiers/priority-transfer-rollback.js +53 -0
- package/benchmark/auto-resolve/fixtures/F32-cli-subscription-renewal/NOTES.md +38 -0
- package/benchmark/auto-resolve/fixtures/F32-cli-subscription-renewal/expected.json +57 -0
- package/benchmark/auto-resolve/fixtures/F32-cli-subscription-renewal/metadata.json +10 -0
- package/benchmark/auto-resolve/fixtures/F32-cli-subscription-renewal/setup.sh +2 -0
- package/benchmark/auto-resolve/fixtures/F32-cli-subscription-renewal/spec.md +70 -0
- package/benchmark/auto-resolve/fixtures/F32-cli-subscription-renewal/task.txt +3 -0
- package/benchmark/auto-resolve/fixtures/F32-cli-subscription-renewal/verifiers/duplicate-renewal-error.js +42 -0
- package/benchmark/auto-resolve/fixtures/F32-cli-subscription-renewal/verifiers/priority-credit-rollback.js +70 -0
- package/benchmark/auto-resolve/fixtures/F4-web-browser-design/NOTES.md +10 -3
- package/benchmark/auto-resolve/fixtures/F5-fix-loop-red-green/NOTES.md +7 -0
- package/benchmark/auto-resolve/fixtures/F6-dep-audit-native-module/NOTES.md +5 -0
- package/benchmark/auto-resolve/fixtures/F7-out-of-scope-trap/NOTES.md +7 -0
- package/benchmark/auto-resolve/fixtures/F8-known-limit-ambiguous/NOTES.md +3 -0
- package/benchmark/auto-resolve/fixtures/F8-known-limit-ambiguous/spec.md +1 -1
- package/benchmark/auto-resolve/fixtures/F9-e2e-ideate-to-resolve/NOTES.md +15 -3
- package/benchmark/auto-resolve/fixtures/F9-e2e-ideate-to-resolve/spec.md +1 -1
- package/benchmark/auto-resolve/fixtures/SCHEMA.md +53 -7
- package/benchmark/auto-resolve/fixtures/retired/F27-cli-subscription-proration/NOTES.md +37 -0
- package/benchmark/auto-resolve/fixtures/retired/F27-cli-subscription-proration/RETIRED.md +13 -0
- package/benchmark/auto-resolve/fixtures/retired/F27-cli-subscription-proration/expected.json +56 -0
- package/benchmark/auto-resolve/fixtures/retired/F27-cli-subscription-proration/metadata.json +10 -0
- package/benchmark/auto-resolve/fixtures/retired/F27-cli-subscription-proration/setup.sh +18 -0
- package/benchmark/auto-resolve/fixtures/retired/F27-cli-subscription-proration/spec.md +69 -0
- package/benchmark/auto-resolve/fixtures/retired/F27-cli-subscription-proration/task.txt +7 -0
- package/benchmark/auto-resolve/fixtures/retired/F27-cli-subscription-proration/verifiers/exact-proration.js +48 -0
- package/benchmark/auto-resolve/fixtures/retired/F27-cli-subscription-proration/verifiers/rules-source-and-conflict.js +79 -0
- package/benchmark/auto-resolve/fixtures/retired/F28-cli-return-authorization/NOTES.md +54 -0
- package/benchmark/auto-resolve/fixtures/retired/F28-cli-return-authorization/RETIRED.md +7 -0
- package/benchmark/auto-resolve/fixtures/retired/F28-cli-return-authorization/expected.json +67 -0
- package/benchmark/auto-resolve/fixtures/retired/F28-cli-return-authorization/metadata.json +10 -0
- package/benchmark/auto-resolve/fixtures/retired/F28-cli-return-authorization/setup.sh +2 -0
- package/benchmark/auto-resolve/fixtures/retired/F28-cli-return-authorization/spec.md +67 -0
- package/benchmark/auto-resolve/fixtures/retired/F28-cli-return-authorization/task.txt +5 -0
- package/benchmark/auto-resolve/fixtures/retired/F28-cli-return-authorization/verifiers/policy-precedence.js +72 -0
- package/benchmark/auto-resolve/fixtures/retired/F28-cli-return-authorization/verifiers/validation-and-immutability.js +43 -0
- package/benchmark/auto-resolve/fixtures/retired/F28-cli-return-authorization/verifiers/validation-boundary.js +116 -0
- package/benchmark/auto-resolve/fixtures/retired/F30-cli-credit-hold-settlement/NOTES.md +35 -0
- package/benchmark/auto-resolve/fixtures/retired/F30-cli-credit-hold-settlement/RETIRED.md +12 -0
- package/benchmark/auto-resolve/fixtures/retired/F30-cli-credit-hold-settlement/expected.json +58 -0
- package/benchmark/auto-resolve/fixtures/retired/F30-cli-credit-hold-settlement/metadata.json +10 -0
- package/benchmark/auto-resolve/fixtures/retired/F30-cli-credit-hold-settlement/setup.sh +2 -0
- package/benchmark/auto-resolve/fixtures/retired/F30-cli-credit-hold-settlement/spec.md +73 -0
- package/benchmark/auto-resolve/fixtures/retired/F30-cli-credit-hold-settlement/task.txt +17 -0
- package/benchmark/auto-resolve/fixtures/retired/F30-cli-credit-hold-settlement/verifiers/mixed-idempotent-settlement.js +53 -0
- package/benchmark/auto-resolve/fixtures/retired/F30-cli-credit-hold-settlement/verifiers/rejection-boundaries.js +74 -0
- package/benchmark/auto-resolve/fixtures/retired/F9-e2e-ideate-to-preflight/NOTES.md +60 -0
- package/benchmark/auto-resolve/fixtures/retired/F9-e2e-ideate-to-preflight/RETIRED.md +29 -0
- package/benchmark/auto-resolve/fixtures/retired/F9-e2e-ideate-to-preflight/expected.json +73 -0
- package/benchmark/auto-resolve/fixtures/retired/F9-e2e-ideate-to-preflight/metadata.json +10 -0
- package/benchmark/auto-resolve/fixtures/retired/F9-e2e-ideate-to-preflight/setup.sh +28 -0
- package/benchmark/auto-resolve/fixtures/retired/F9-e2e-ideate-to-preflight/spec.md +58 -0
- package/benchmark/auto-resolve/fixtures/retired/F9-e2e-ideate-to-preflight/task.txt +5 -0
- package/benchmark/auto-resolve/results/20260510-f16-f23-f25-combined-proof/full-pipeline-pair-gate.json +82 -0
- package/benchmark/auto-resolve/results/20260510-f16-f23-f25-combined-proof/full-pipeline-pair-gate.md +18 -0
- package/benchmark/auto-resolve/results/20260510-f16-f23-f25-combined-proof/headroom-gate.json +46 -0
- package/benchmark/auto-resolve/results/20260510-f16-f23-f25-combined-proof/headroom-gate.md +17 -0
- package/benchmark/auto-resolve/run-real-benchmark.md +303 -0
- package/benchmark/auto-resolve/scripts/audit-headroom-rejections.py +441 -0
- package/benchmark/auto-resolve/scripts/audit-pair-evidence.py +1256 -0
- package/benchmark/auto-resolve/scripts/build-pair-eligible-manifest.py +147 -15
- package/benchmark/auto-resolve/scripts/check-f9-artifacts.py +28 -16
- package/benchmark/auto-resolve/scripts/collect-swebench-predictions.py +11 -1
- package/benchmark/auto-resolve/scripts/compile-report.py +208 -46
- package/benchmark/auto-resolve/scripts/fetch-swebench-instances.py +22 -4
- package/benchmark/auto-resolve/scripts/frozen-verify-gate.py +175 -30
- package/benchmark/auto-resolve/scripts/full-pipeline-pair-gate.py +408 -46
- package/benchmark/auto-resolve/scripts/headroom-gate.py +270 -39
- package/benchmark/auto-resolve/scripts/iter-0033c-compare.py +164 -33
- package/benchmark/auto-resolve/scripts/iter-0033c-l1-summary.py +97 -0
- package/benchmark/auto-resolve/scripts/judge-opus-pass.sh +150 -38
- package/benchmark/auto-resolve/scripts/judge.sh +153 -26
- package/benchmark/auto-resolve/scripts/oracle-scope-tier-a.py +12 -5
- package/benchmark/auto-resolve/scripts/oracle-scope-tier-b.py +25 -2
- package/benchmark/auto-resolve/scripts/pair-candidate-frontier.py +469 -0
- package/benchmark/auto-resolve/scripts/pair-plan-idgen.py +5 -5
- package/benchmark/auto-resolve/scripts/pair-plan-lint.py +9 -2
- package/benchmark/auto-resolve/scripts/pair-rejected-fixtures.sh +91 -0
- package/benchmark/auto-resolve/scripts/pair_evidence_contract.py +269 -0
- package/benchmark/auto-resolve/scripts/prepare-swebench-frozen-case.py +39 -10
- package/benchmark/auto-resolve/scripts/prepare-swebench-frozen-corpus.py +34 -4
- package/benchmark/auto-resolve/scripts/prepare-swebench-solver-worktree.py +23 -5
- package/benchmark/auto-resolve/scripts/recent-benchmark-summary.py +232 -0
- package/benchmark/auto-resolve/scripts/run-fixture.sh +118 -51
- package/benchmark/auto-resolve/scripts/run-frozen-verify-pair.sh +211 -39
- package/benchmark/auto-resolve/scripts/run-full-pipeline-pair-candidate.sh +335 -39
- package/benchmark/auto-resolve/scripts/run-headroom-candidate.sh +249 -6
- package/benchmark/auto-resolve/scripts/run-iter-0033c.sh +22 -48
- package/benchmark/auto-resolve/scripts/run-suite.sh +44 -7
- package/benchmark/auto-resolve/scripts/run-swebench-frozen-corpus.sh +120 -19
- package/benchmark/auto-resolve/scripts/run-swebench-solver-batch.sh +32 -14
- package/benchmark/auto-resolve/scripts/ship-gate.py +219 -50
- package/benchmark/auto-resolve/scripts/solo-ceiling-avoidance.py +53 -0
- package/benchmark/auto-resolve/scripts/solo-headroom-hypothesis.py +77 -0
- package/benchmark/auto-resolve/scripts/swebench-frozen-matrix.py +239 -26
- package/benchmark/auto-resolve/scripts/test-audit-headroom-rejections.sh +288 -0
- package/benchmark/auto-resolve/scripts/test-audit-pair-evidence.sh +1672 -0
- package/benchmark/auto-resolve/scripts/test-benchmark-arg-parsing.sh +933 -0
- package/benchmark/auto-resolve/scripts/test-build-pair-eligible-manifest.sh +491 -0
- package/benchmark/auto-resolve/scripts/test-check-f9-artifacts.sh +91 -0
- package/benchmark/auto-resolve/scripts/test-frozen-verify-gate.sh +328 -3
- package/benchmark/auto-resolve/scripts/test-full-pipeline-pair-gate.sh +497 -18
- package/benchmark/auto-resolve/scripts/test-headroom-gate.sh +331 -14
- package/benchmark/auto-resolve/scripts/test-iter-0033c-compare.sh +525 -0
- package/benchmark/auto-resolve/scripts/test-iter-0033c-l1-summary.sh +254 -0
- package/benchmark/auto-resolve/scripts/test-lint-fixtures.sh +580 -0
- package/benchmark/auto-resolve/scripts/test-pair-candidate-frontier.sh +591 -0
- package/benchmark/auto-resolve/scripts/test-run-full-pipeline-pair-candidate.sh +497 -0
- package/benchmark/auto-resolve/scripts/test-run-headroom-candidate.sh +401 -0
- package/benchmark/auto-resolve/scripts/test-run-swebench-solver-batch.sh +111 -0
- package/benchmark/auto-resolve/scripts/test-ship-gate.sh +1189 -0
- package/benchmark/auto-resolve/scripts/test-swebench-frozen-case.sh +924 -5
- package/benchmark/auto-resolve/shadow-fixtures/S1-cli-lang-flag/NOTES.md +28 -0
- package/benchmark/auto-resolve/shadow-fixtures/S1-cli-lang-flag/expected.json +63 -0
- package/benchmark/auto-resolve/shadow-fixtures/S1-cli-lang-flag/metadata.json +10 -0
- package/benchmark/auto-resolve/shadow-fixtures/S1-cli-lang-flag/setup.sh +3 -0
- package/benchmark/auto-resolve/shadow-fixtures/S1-cli-lang-flag/spec.md +47 -0
- package/benchmark/auto-resolve/shadow-fixtures/S1-cli-lang-flag/task.txt +1 -0
- package/benchmark/auto-resolve/shadow-fixtures/S2-cli-inventory-reservation/NOTES.md +34 -0
- package/benchmark/auto-resolve/shadow-fixtures/S2-cli-inventory-reservation/expected.json +53 -0
- package/benchmark/auto-resolve/shadow-fixtures/S2-cli-inventory-reservation/metadata.json +10 -0
- package/benchmark/auto-resolve/shadow-fixtures/S2-cli-inventory-reservation/setup.sh +3 -0
- package/benchmark/auto-resolve/shadow-fixtures/S2-cli-inventory-reservation/spec.md +50 -0
- package/benchmark/auto-resolve/shadow-fixtures/S2-cli-inventory-reservation/task.txt +1 -0
- package/benchmark/auto-resolve/shadow-fixtures/S2-cli-inventory-reservation/verifiers/duplicate-order-error.js +27 -0
- package/benchmark/auto-resolve/shadow-fixtures/S2-cli-inventory-reservation/verifiers/priority-stock-reservation.js +44 -0
- package/benchmark/auto-resolve/shadow-fixtures/S3-cli-ticket-assignment/NOTES.md +34 -0
- package/benchmark/auto-resolve/shadow-fixtures/S3-cli-ticket-assignment/expected.json +55 -0
- package/benchmark/auto-resolve/shadow-fixtures/S3-cli-ticket-assignment/metadata.json +10 -0
- package/benchmark/auto-resolve/shadow-fixtures/S3-cli-ticket-assignment/setup.sh +3 -0
- package/benchmark/auto-resolve/shadow-fixtures/S3-cli-ticket-assignment/spec.md +52 -0
- package/benchmark/auto-resolve/shadow-fixtures/S3-cli-ticket-assignment/task.txt +1 -0
- package/benchmark/auto-resolve/shadow-fixtures/S3-cli-ticket-assignment/verifiers/duplicate-ticket-error.js +29 -0
- package/benchmark/auto-resolve/shadow-fixtures/S3-cli-ticket-assignment/verifiers/priority-agent-assignment.js +48 -0
- package/benchmark/auto-resolve/shadow-fixtures/S4-cli-return-routing/NOTES.md +34 -0
- package/benchmark/auto-resolve/shadow-fixtures/S4-cli-return-routing/expected.json +55 -0
- package/benchmark/auto-resolve/shadow-fixtures/S4-cli-return-routing/metadata.json +10 -0
- package/benchmark/auto-resolve/shadow-fixtures/S4-cli-return-routing/setup.sh +3 -0
- package/benchmark/auto-resolve/shadow-fixtures/S4-cli-return-routing/spec.md +55 -0
- package/benchmark/auto-resolve/shadow-fixtures/S4-cli-return-routing/task.txt +1 -0
- package/benchmark/auto-resolve/shadow-fixtures/S4-cli-return-routing/verifiers/duplicate-return-error.js +43 -0
- package/benchmark/auto-resolve/shadow-fixtures/S4-cli-return-routing/verifiers/priority-return-routing.js +70 -0
- package/benchmark/auto-resolve/shadow-fixtures/S5-cli-credit-grant-ledger/NOTES.md +37 -0
- package/benchmark/auto-resolve/shadow-fixtures/S5-cli-credit-grant-ledger/expected.json +54 -0
- package/benchmark/auto-resolve/shadow-fixtures/S5-cli-credit-grant-ledger/metadata.json +10 -0
- package/benchmark/auto-resolve/shadow-fixtures/S5-cli-credit-grant-ledger/setup.sh +3 -0
- package/benchmark/auto-resolve/shadow-fixtures/S5-cli-credit-grant-ledger/spec.md +59 -0
- package/benchmark/auto-resolve/shadow-fixtures/S5-cli-credit-grant-ledger/task.txt +1 -0
- package/benchmark/auto-resolve/shadow-fixtures/S5-cli-credit-grant-ledger/verifiers/credit-ledger-priority.js +98 -0
- package/benchmark/auto-resolve/shadow-fixtures/S5-cli-credit-grant-ledger/verifiers/duplicate-charge-error.js +38 -0
- package/benchmark/auto-resolve/shadow-fixtures/S6-cli-refund-window-ledger/NOTES.md +36 -0
- package/benchmark/auto-resolve/shadow-fixtures/S6-cli-refund-window-ledger/expected.json +56 -0
- package/benchmark/auto-resolve/shadow-fixtures/S6-cli-refund-window-ledger/metadata.json +10 -0
- package/benchmark/auto-resolve/shadow-fixtures/S6-cli-refund-window-ledger/setup.sh +3 -0
- package/benchmark/auto-resolve/shadow-fixtures/S6-cli-refund-window-ledger/spec.md +59 -0
- package/benchmark/auto-resolve/shadow-fixtures/S6-cli-refund-window-ledger/task.txt +1 -0
- package/benchmark/auto-resolve/shadow-fixtures/S6-cli-refund-window-ledger/verifiers/duplicate-refund-error.js +41 -0
- package/benchmark/auto-resolve/shadow-fixtures/S6-cli-refund-window-ledger/verifiers/priority-refund-ledger.js +65 -0
- package/bin/devlyn.js +221 -17
- package/config/skills/_shared/adapters/README.md +3 -0
- package/config/skills/_shared/adapters/gpt-5-5.md +5 -1
- package/config/skills/_shared/adapters/opus-4-7.md +9 -1
- package/config/skills/_shared/archive_run.py +78 -6
- package/config/skills/_shared/codex-config.md +5 -4
- package/config/skills/_shared/codex-monitored.sh +46 -1
- package/config/skills/_shared/collect-codex-findings.py +20 -5
- package/config/skills/_shared/engine-preflight.md +17 -13
- package/config/skills/_shared/runtime-principles.md +6 -9
- package/config/skills/_shared/spec-verify-check.py +2664 -107
- package/config/skills/_shared/verify-merge-findings.py +1369 -19
- package/config/skills/devlyn:design-ui/SKILL.md +364 -0
- package/config/skills/devlyn:ideate/SKILL.md +7 -4
- package/config/skills/devlyn:ideate/references/elicitation.md +50 -4
- package/config/skills/devlyn:ideate/references/from-spec-mode.md +26 -4
- package/config/skills/devlyn:ideate/references/project-mode.md +20 -1
- package/config/skills/devlyn:ideate/references/spec-template.md +10 -1
- package/config/skills/devlyn:resolve/SKILL.md +78 -26
- package/config/skills/devlyn:resolve/references/free-form-mode.md +15 -0
- package/config/skills/devlyn:resolve/references/phases/build-gate.md +2 -2
- package/config/skills/devlyn:resolve/references/phases/implement.md +1 -1
- package/config/skills/devlyn:resolve/references/phases/probe-derive.md +74 -2
- package/config/skills/devlyn:resolve/references/phases/verify.md +80 -29
- package/config/skills/devlyn:resolve/references/state-schema.md +9 -4
- package/package.json +47 -2
- package/scripts/lint-fixtures.sh +349 -0
- package/scripts/lint-shadow-fixtures.sh +58 -0
- package/scripts/lint-skills.sh +3645 -95
|
@@ -2,8 +2,9 @@
|
|
|
2
2
|
# run-full-pipeline-pair-candidate.sh — measure full-pipeline L2/pair candidates.
|
|
3
3
|
#
|
|
4
4
|
# Runs bare + solo_claude first and applies headroom-gate.py. Only if the set
|
|
5
|
-
# leaves room for L2 does it run
|
|
6
|
-
# full-pipeline-pair-gate.py.
|
|
5
|
+
# leaves room for L2 does it run the selected pair arm, rejudge, and apply
|
|
6
|
+
# full-pipeline-pair-gate.py. Default pair arm is l2_risk_probes because that is
|
|
7
|
+
# the current measured solo<pair proof path.
|
|
7
8
|
|
|
8
9
|
set -euo pipefail
|
|
9
10
|
|
|
@@ -16,43 +17,120 @@ Options:
|
|
|
16
17
|
--run-id ID
|
|
17
18
|
--bare-max N
|
|
18
19
|
--solo-max N
|
|
20
|
+
--min-bare-headroom N
|
|
21
|
+
--min-solo-headroom N
|
|
19
22
|
--min-fixtures N
|
|
20
23
|
--min-pair-margin N
|
|
21
|
-
--max-pair-solo-wall-ratio N
|
|
22
|
-
--pair-arm ARM
|
|
24
|
+
--max-pair-solo-wall-ratio N (default: 3)
|
|
25
|
+
--pair-arm ARM (default: l2_risk_probes; use l2_gated only for diagnostics)
|
|
23
26
|
--reuse-calibrated-from RUN_ID
|
|
27
|
+
--allow-rejected-fixtures
|
|
28
|
+
allow rejected/ceiling fixtures for diagnostics only
|
|
29
|
+
--dry-run validate args/fixtures and print replay command only
|
|
24
30
|
EOF
|
|
25
31
|
exit "$code"
|
|
26
32
|
}
|
|
27
33
|
|
|
34
|
+
require_value() {
|
|
35
|
+
local flag="$1"
|
|
36
|
+
local value="${2:-}"
|
|
37
|
+
if [ -z "$value" ] || [[ "$value" == --* ]]; then
|
|
38
|
+
echo "$flag requires a value" >&2
|
|
39
|
+
exit 1
|
|
40
|
+
fi
|
|
41
|
+
}
|
|
42
|
+
|
|
28
43
|
RUN_ID=""
|
|
29
44
|
BARE_MAX=60
|
|
30
45
|
SOLO_MAX=80
|
|
46
|
+
MIN_BARE_HEADROOM=5
|
|
47
|
+
MIN_SOLO_HEADROOM=5
|
|
31
48
|
MIN_FIXTURES=2
|
|
32
49
|
MIN_PAIR_MARGIN=5
|
|
33
|
-
MAX_PAIR_SOLO_WALL_RATIO=
|
|
34
|
-
PAIR_ARM="
|
|
50
|
+
MAX_PAIR_SOLO_WALL_RATIO=3
|
|
51
|
+
PAIR_ARM="l2_risk_probes"
|
|
35
52
|
REUSE_CALIBRATED_FROM=""
|
|
53
|
+
ALLOW_REJECTED_FIXTURES=0
|
|
54
|
+
DRY_RUN=0
|
|
36
55
|
FIXTURES=()
|
|
37
56
|
while [ $# -gt 0 ]; do
|
|
38
57
|
case "$1" in
|
|
39
|
-
--run-id) RUN_ID="$2"; shift 2;;
|
|
40
|
-
--bare-max) BARE_MAX="$2"; shift 2;;
|
|
41
|
-
--solo-max) SOLO_MAX="$2"; shift 2;;
|
|
42
|
-
--min-
|
|
43
|
-
--min-
|
|
44
|
-
--
|
|
45
|
-
--pair-
|
|
46
|
-
--
|
|
58
|
+
--run-id) require_value "$1" "${2:-}"; RUN_ID="$2"; shift 2;;
|
|
59
|
+
--bare-max) require_value "$1" "${2:-}"; BARE_MAX="$2"; shift 2;;
|
|
60
|
+
--solo-max) require_value "$1" "${2:-}"; SOLO_MAX="$2"; shift 2;;
|
|
61
|
+
--min-bare-headroom) require_value "$1" "${2:-}"; MIN_BARE_HEADROOM="$2"; shift 2;;
|
|
62
|
+
--min-solo-headroom) require_value "$1" "${2:-}"; MIN_SOLO_HEADROOM="$2"; shift 2;;
|
|
63
|
+
--min-fixtures) require_value "$1" "${2:-}"; MIN_FIXTURES="$2"; shift 2;;
|
|
64
|
+
--min-pair-margin) require_value "$1" "${2:-}"; MIN_PAIR_MARGIN="$2"; shift 2;;
|
|
65
|
+
--max-pair-solo-wall-ratio) require_value "$1" "${2:-}"; MAX_PAIR_SOLO_WALL_RATIO="$2"; shift 2;;
|
|
66
|
+
--pair-arm) require_value "$1" "${2:-}"; PAIR_ARM="$2"; shift 2;;
|
|
67
|
+
--reuse-calibrated-from) require_value "$1" "${2:-}"; REUSE_CALIBRATED_FROM="$2"; shift 2;;
|
|
68
|
+
--allow-rejected-fixtures) ALLOW_REJECTED_FIXTURES=1; shift;;
|
|
69
|
+
--dry-run) DRY_RUN=1; shift;;
|
|
47
70
|
-h|--help) usage 0;;
|
|
48
|
-
|
|
71
|
+
[FS][0-9]*) FIXTURES+=("$1"); shift;;
|
|
49
72
|
*) echo "unknown arg: $1" >&2; usage;;
|
|
50
73
|
esac
|
|
51
74
|
done
|
|
75
|
+
|
|
76
|
+
for threshold in BARE_MAX SOLO_MAX MIN_BARE_HEADROOM MIN_SOLO_HEADROOM MIN_FIXTURES MIN_PAIR_MARGIN; do
|
|
77
|
+
value="${!threshold}"
|
|
78
|
+
case "$threshold" in
|
|
79
|
+
BARE_MAX) flag="bare-max" ;;
|
|
80
|
+
SOLO_MAX) flag="solo-max" ;;
|
|
81
|
+
MIN_BARE_HEADROOM) flag="min-bare-headroom" ;;
|
|
82
|
+
MIN_SOLO_HEADROOM) flag="min-solo-headroom" ;;
|
|
83
|
+
MIN_FIXTURES) flag="min-fixtures" ;;
|
|
84
|
+
MIN_PAIR_MARGIN) flag="min-pair-margin" ;;
|
|
85
|
+
esac
|
|
86
|
+
if [[ ! "$value" =~ ^[0-9]+$ ]]; then
|
|
87
|
+
echo "--$flag must be an integer: $value" >&2
|
|
88
|
+
exit 1
|
|
89
|
+
fi
|
|
90
|
+
done
|
|
91
|
+
if [ "$MIN_FIXTURES" -lt 1 ]; then
|
|
92
|
+
echo "--min-fixtures must be >= 1" >&2
|
|
93
|
+
exit 1
|
|
94
|
+
fi
|
|
95
|
+
if [ "$MIN_BARE_HEADROOM" -lt 0 ]; then
|
|
96
|
+
echo "--min-bare-headroom must be >= 0" >&2
|
|
97
|
+
exit 1
|
|
98
|
+
fi
|
|
99
|
+
if [ "$MIN_SOLO_HEADROOM" -lt 0 ]; then
|
|
100
|
+
echo "--min-solo-headroom must be >= 0" >&2
|
|
101
|
+
exit 1
|
|
102
|
+
fi
|
|
103
|
+
if [ -n "$MAX_PAIR_SOLO_WALL_RATIO" ]; then
|
|
104
|
+
if ! [[ "$MAX_PAIR_SOLO_WALL_RATIO" =~ ^[0-9]+([.][0-9]+)?$ ]]; then
|
|
105
|
+
echo "--max-pair-solo-wall-ratio must be a positive number: $MAX_PAIR_SOLO_WALL_RATIO" >&2
|
|
106
|
+
exit 1
|
|
107
|
+
fi
|
|
108
|
+
if ! awk "BEGIN { exit !($MAX_PAIR_SOLO_WALL_RATIO > 0) }"; then
|
|
109
|
+
echo "--max-pair-solo-wall-ratio must be > 0" >&2
|
|
110
|
+
exit 1
|
|
111
|
+
fi
|
|
112
|
+
fi
|
|
52
113
|
[ ${#FIXTURES[@]} -gt 0 ] || usage
|
|
53
114
|
|
|
115
|
+
case "$PAIR_ARM" in
|
|
116
|
+
l2_risk_probes|l2_gated) ;;
|
|
117
|
+
l2_forced)
|
|
118
|
+
echo "pair-arm l2_forced is retired: it leaks pair-awareness before IMPLEMENT; use l2_risk_probes for current proof runs or l2_gated for diagnostics." >&2
|
|
119
|
+
exit 1
|
|
120
|
+
;;
|
|
121
|
+
*)
|
|
122
|
+
echo "pair-arm must be l2_risk_probes or l2_gated (diagnostic): $PAIR_ARM" >&2
|
|
123
|
+
exit 1
|
|
124
|
+
;;
|
|
125
|
+
esac
|
|
126
|
+
|
|
54
127
|
BENCH_ROOT="$(cd "$(dirname "$0")/.." && pwd)"
|
|
55
128
|
REPO_ROOT="$(cd "$BENCH_ROOT/../.." && pwd)"
|
|
129
|
+
source "$BENCH_ROOT/scripts/pair-rejected-fixtures.sh"
|
|
130
|
+
if ! declare -F rejected_pair_fixture_reason >/dev/null; then
|
|
131
|
+
echo "rejected fixture registry must define rejected_pair_fixture_reason" >&2
|
|
132
|
+
exit 1
|
|
133
|
+
fi
|
|
56
134
|
|
|
57
135
|
if [ -z "$RUN_ID" ]; then
|
|
58
136
|
TS=$(date -u +%Y%m%dT%H%M%SZ)
|
|
@@ -60,50 +138,245 @@ if [ -z "$RUN_ID" ]; then
|
|
|
60
138
|
RUN_ID="${TS}-${SHA}-full-pipeline-pair"
|
|
61
139
|
fi
|
|
62
140
|
|
|
141
|
+
print_command() {
|
|
142
|
+
local cmd
|
|
143
|
+
if [ "${DEVLYN_BENCHMARK_CLI_SUBCOMMAND:-}" = "pair" ]; then
|
|
144
|
+
cmd=(npx devlyn-cli benchmark pair --run-id "$RUN_ID")
|
|
145
|
+
else
|
|
146
|
+
cmd=(bash "$0" --run-id "$RUN_ID")
|
|
147
|
+
fi
|
|
148
|
+
cmd+=(--bare-max "$BARE_MAX")
|
|
149
|
+
cmd+=(--solo-max "$SOLO_MAX")
|
|
150
|
+
cmd+=(--min-bare-headroom "$MIN_BARE_HEADROOM")
|
|
151
|
+
cmd+=(--min-solo-headroom "$MIN_SOLO_HEADROOM")
|
|
152
|
+
cmd+=(--min-fixtures "$MIN_FIXTURES")
|
|
153
|
+
cmd+=(--min-pair-margin "$MIN_PAIR_MARGIN")
|
|
154
|
+
[ -z "$MAX_PAIR_SOLO_WALL_RATIO" ] || cmd+=(--max-pair-solo-wall-ratio "$MAX_PAIR_SOLO_WALL_RATIO")
|
|
155
|
+
cmd+=(--pair-arm "$PAIR_ARM")
|
|
156
|
+
[ -z "$REUSE_CALIBRATED_FROM" ] || cmd+=(--reuse-calibrated-from "$REUSE_CALIBRATED_FROM")
|
|
157
|
+
[ "$ALLOW_REJECTED_FIXTURES" -eq 0 ] || cmd+=(--allow-rejected-fixtures)
|
|
158
|
+
[ "$DRY_RUN" -eq 0 ] || cmd+=(--dry-run)
|
|
159
|
+
cmd+=("${FIXTURES[@]}")
|
|
160
|
+
printf 'Command: '
|
|
161
|
+
printf '%q ' "${cmd[@]}"
|
|
162
|
+
printf '\n'
|
|
163
|
+
}
|
|
164
|
+
|
|
165
|
+
run_gate_with_report() {
|
|
166
|
+
local label="$1"
|
|
167
|
+
local report="$2"
|
|
168
|
+
shift 2
|
|
169
|
+
set +e
|
|
170
|
+
"$@"
|
|
171
|
+
local status=$?
|
|
172
|
+
set -e
|
|
173
|
+
if [ -f "$report" ]; then
|
|
174
|
+
cat "$report"
|
|
175
|
+
else
|
|
176
|
+
echo "[full-pipeline-pair] ${label} report missing: $report" >&2
|
|
177
|
+
fi
|
|
178
|
+
return "$status"
|
|
179
|
+
}
|
|
180
|
+
|
|
181
|
+
fixture_exists() {
|
|
182
|
+
local fid="$1"
|
|
183
|
+
[ -d "$BENCH_ROOT/fixtures/$fid" ] || [ -d "$BENCH_ROOT/shadow-fixtures/$fid" ]
|
|
184
|
+
}
|
|
185
|
+
|
|
186
|
+
fixture_dir() {
|
|
187
|
+
local fid="$1"
|
|
188
|
+
if [ -d "$BENCH_ROOT/fixtures/$fid" ]; then
|
|
189
|
+
printf '%s\n' "$BENCH_ROOT/fixtures/$fid"
|
|
190
|
+
else
|
|
191
|
+
printf '%s\n' "$BENCH_ROOT/shadow-fixtures/$fid"
|
|
192
|
+
fi
|
|
193
|
+
}
|
|
194
|
+
|
|
195
|
+
is_shadow_fixture() {
|
|
196
|
+
local fid="$1"
|
|
197
|
+
[ -d "$BENCH_ROOT/shadow-fixtures/$fid" ]
|
|
198
|
+
}
|
|
199
|
+
|
|
200
|
+
retired_fixture_exists() {
|
|
201
|
+
local fid="$1"
|
|
202
|
+
[ -d "$BENCH_ROOT/fixtures/retired/$fid" ]
|
|
203
|
+
}
|
|
204
|
+
|
|
205
|
+
fixture_smoke_only() {
|
|
206
|
+
local fid="$1"
|
|
207
|
+
[[ "$fid" == S1 || "$fid" == S1-* ]]
|
|
208
|
+
}
|
|
209
|
+
|
|
210
|
+
fixture_category() {
|
|
211
|
+
local dir="$1"
|
|
212
|
+
python3 - "$dir/metadata.json" <<'PY'
|
|
213
|
+
import json
|
|
214
|
+
import sys
|
|
215
|
+
|
|
216
|
+
try:
|
|
217
|
+
with open(sys.argv[1], encoding="utf-8") as handle:
|
|
218
|
+
print(json.load(handle).get("category", ""))
|
|
219
|
+
except FileNotFoundError:
|
|
220
|
+
print("")
|
|
221
|
+
PY
|
|
222
|
+
}
|
|
223
|
+
|
|
224
|
+
fixture_has_solo_headroom_hypothesis() {
|
|
225
|
+
local dir="$1"
|
|
226
|
+
python3 "$BENCH_ROOT/scripts/solo-headroom-hypothesis.py" --expected-json "$dir/expected.json" "$dir/spec.md"
|
|
227
|
+
}
|
|
228
|
+
|
|
229
|
+
fixture_has_solo_ceiling_avoidance_note() {
|
|
230
|
+
local dir="$1"
|
|
231
|
+
python3 "$BENCH_ROOT/scripts/solo-ceiling-avoidance.py" "$dir/NOTES.md"
|
|
232
|
+
}
|
|
233
|
+
|
|
234
|
+
fixture_has_pair_evidence() {
|
|
235
|
+
local fid="$1"
|
|
236
|
+
python3 - "$BENCH_ROOT/results" "$fid" <<'PY'
|
|
237
|
+
import json
|
|
238
|
+
import pathlib
|
|
239
|
+
import sys
|
|
240
|
+
|
|
241
|
+
results = pathlib.Path(sys.argv[1])
|
|
242
|
+
fixture = sys.argv[2]
|
|
243
|
+
if not results.is_dir():
|
|
244
|
+
sys.exit(1)
|
|
245
|
+
for path in results.glob("*/full-pipeline-pair-gate.json"):
|
|
246
|
+
try:
|
|
247
|
+
data = json.loads(path.read_text(encoding="utf-8"))
|
|
248
|
+
except (OSError, json.JSONDecodeError):
|
|
249
|
+
continue
|
|
250
|
+
if data.get("verdict") != "PASS":
|
|
251
|
+
continue
|
|
252
|
+
rows = data.get("rows")
|
|
253
|
+
if not isinstance(rows, list):
|
|
254
|
+
continue
|
|
255
|
+
for row in rows:
|
|
256
|
+
if isinstance(row, dict) and row.get("fixture") == fixture and row.get("status") == "PASS":
|
|
257
|
+
sys.exit(0)
|
|
258
|
+
sys.exit(1)
|
|
259
|
+
PY
|
|
260
|
+
}
|
|
261
|
+
|
|
262
|
+
validate_fixtures() {
|
|
263
|
+
local missing=0
|
|
264
|
+
local fid reason dir category
|
|
265
|
+
for fid in "${FIXTURES[@]}"; do
|
|
266
|
+
if ! fixture_exists "$fid"; then
|
|
267
|
+
if retired_fixture_exists "$fid"; then
|
|
268
|
+
echo "fixture is retired and is not rerun by pair-candidate runners: $fid. Use preserved results/docs for historical replay." >&2
|
|
269
|
+
missing=1
|
|
270
|
+
continue
|
|
271
|
+
fi
|
|
272
|
+
echo "fixture not found in fixtures/ or shadow-fixtures/: $fid" >&2
|
|
273
|
+
missing=1
|
|
274
|
+
continue
|
|
275
|
+
fi
|
|
276
|
+
if [ "$DRY_RUN" -eq 0 ] && fixture_smoke_only "$fid"; then
|
|
277
|
+
echo "fixture is smoke-only and cannot run providers: $fid. Use --dry-run for runner/package validation." >&2
|
|
278
|
+
missing=1
|
|
279
|
+
continue
|
|
280
|
+
fi
|
|
281
|
+
reason="$(rejected_pair_fixture_reason "$fid" || true)"
|
|
282
|
+
if [ "$ALLOW_REJECTED_FIXTURES" -eq 0 ]; then
|
|
283
|
+
if [ -n "$reason" ]; then
|
|
284
|
+
echo "fixture rejected for pair-candidate runs: $fid ($reason). Use --allow-rejected-fixtures for diagnostics only." >&2
|
|
285
|
+
missing=1
|
|
286
|
+
continue
|
|
287
|
+
fi
|
|
288
|
+
fi
|
|
289
|
+
if [ -z "$reason" ]; then
|
|
290
|
+
dir="$(fixture_dir "$fid")"
|
|
291
|
+
category="$(fixture_category "$dir")"
|
|
292
|
+
if [ "$category" = "high-risk" ] && ! fixture_has_pair_evidence "$fid"; then
|
|
293
|
+
if ! fixture_has_solo_headroom_hypothesis "$dir"; then
|
|
294
|
+
echo "fixture spec.md needs a solo-headroom hypothesis with solo_claude miss and observable command from expected.json before provider spend: $fid" >&2
|
|
295
|
+
missing=1
|
|
296
|
+
fi
|
|
297
|
+
if is_shadow_fixture "$fid" && ! fixture_has_solo_ceiling_avoidance_note "$dir"; then
|
|
298
|
+
echo "shadow fixture NOTES.md needs ## Solo ceiling avoidance with solo_claude, a rejected/solo-saturated control comparison, and headroom reasoning before provider spend: $fid" >&2
|
|
299
|
+
missing=1
|
|
300
|
+
fi
|
|
301
|
+
fi
|
|
302
|
+
fi
|
|
303
|
+
done
|
|
304
|
+
[ "$missing" -eq 0 ] || exit 1
|
|
305
|
+
}
|
|
306
|
+
|
|
63
307
|
echo ""
|
|
64
308
|
echo "═══ Full-Pipeline Pair Candidate Run ═══"
|
|
65
309
|
echo "Run-id: $RUN_ID"
|
|
66
310
|
echo "Fixtures: ${FIXTURES[*]}"
|
|
67
311
|
echo "Arms: bare solo_claude $PAIR_ARM"
|
|
312
|
+
echo "Headroom: bare <= $BARE_MAX (headroom >= $MIN_BARE_HEADROOM), solo_claude <= $SOLO_MAX (headroom >= $MIN_SOLO_HEADROOM), baseline evidence-complete, min fixtures $MIN_FIXTURES"
|
|
313
|
+
echo "Pair: $PAIR_ARM evidence-clean, canonical trigger, margin >= +$MIN_PAIR_MARGIN${MAX_PAIR_SOLO_WALL_RATIO:+, wall ratio <= $MAX_PAIR_SOLO_WALL_RATIO}"
|
|
68
314
|
[ -z "$REUSE_CALIBRATED_FROM" ] || echo "Reuse: bare+solo from $REUSE_CALIBRATED_FROM"
|
|
315
|
+
[ "$DRY_RUN" -eq 0 ] || echo "Mode: DRY RUN (no model/provider invocations)"
|
|
316
|
+
print_command
|
|
69
317
|
echo ""
|
|
70
318
|
|
|
71
|
-
|
|
72
|
-
|
|
73
|
-
|
|
74
|
-
|
|
75
|
-
|
|
76
|
-
|
|
77
|
-
|
|
78
|
-
|
|
79
|
-
|
|
80
|
-
|
|
81
|
-
|
|
82
|
-
|
|
83
|
-
|
|
84
|
-
|
|
85
|
-
|
|
86
|
-
|
|
87
|
-
|
|
88
|
-
|
|
89
|
-
|
|
319
|
+
validate_fixtures
|
|
320
|
+
|
|
321
|
+
if [ "$DRY_RUN" -eq 1 ] && [ "${#FIXTURES[@]}" -lt "$MIN_FIXTURES" ]; then
|
|
322
|
+
echo "[full-pipeline-pair] DRY RUN failed — ${#FIXTURES[@]} fixture(s) supplied, --min-fixtures requires $MIN_FIXTURES." >&2
|
|
323
|
+
exit 1
|
|
324
|
+
fi
|
|
325
|
+
|
|
326
|
+
if [ "$DRY_RUN" -eq 1 ]; then
|
|
327
|
+
echo "[full-pipeline-pair] DRY RUN complete — fixtures resolved, no arms or judges executed."
|
|
328
|
+
exit 0
|
|
329
|
+
fi
|
|
330
|
+
|
|
331
|
+
mirror_skills() {
|
|
332
|
+
local src_skills="$REPO_ROOT/config/skills"
|
|
333
|
+
local dst_skills="$REPO_ROOT/.claude/skills"
|
|
334
|
+
local mirrored=0
|
|
335
|
+
local src_dir name staging
|
|
336
|
+
mkdir -p "$dst_skills"
|
|
337
|
+
for src_dir in "$src_skills"/*/; do
|
|
338
|
+
[ -d "$src_dir" ] || continue
|
|
339
|
+
name=$(basename "$src_dir")
|
|
340
|
+
case "$name" in
|
|
341
|
+
devlyn:auto-resolve-workspace|devlyn:ideate-workspace|preflight-workspace|roadmap-archival-workspace)
|
|
342
|
+
continue ;;
|
|
343
|
+
esac
|
|
344
|
+
staging="$dst_skills/.${name}.staging"
|
|
345
|
+
rm -rf "$staging"
|
|
346
|
+
cp -R "$src_dir" "$staging"
|
|
347
|
+
rm -rf "$dst_skills/$name"
|
|
348
|
+
mv "$staging" "$dst_skills/$name"
|
|
349
|
+
mirrored=$((mirrored + 1))
|
|
350
|
+
done
|
|
351
|
+
echo "[full-pipeline-pair] mirrored $mirrored committed skill(s): config/skills/ -> .claude/skills/"
|
|
352
|
+
}
|
|
90
353
|
|
|
91
354
|
copy_calibrated_arm() {
|
|
92
355
|
local fid="$1"
|
|
93
356
|
local arm="$2"
|
|
94
357
|
local src="$BENCH_ROOT/results/$REUSE_CALIBRATED_FROM/$fid/$arm"
|
|
95
358
|
local dst="$BENCH_ROOT/results/$RUN_ID/$fid/$arm"
|
|
96
|
-
if [ -
|
|
359
|
+
if [ -e "$dst" ]; then
|
|
360
|
+
[ -d "$dst" ] || { echo "reuse destination is not a directory: $dst" >&2; exit 1; }
|
|
361
|
+
for required in result.json verify.json diff.patch; do
|
|
362
|
+
[ -f "$dst/$required" ] || { echo "reuse destination incomplete $required: $dst" >&2; exit 1; }
|
|
363
|
+
done
|
|
97
364
|
echo "[full-pipeline-pair] reuse skip: $fid / $arm already exists in $RUN_ID"
|
|
98
365
|
return 0
|
|
99
366
|
fi
|
|
100
367
|
[ -d "$src" ] || { echo "reuse source missing: $src" >&2; exit 1; }
|
|
101
|
-
|
|
368
|
+
for required in result.json verify.json diff.patch; do
|
|
369
|
+
[ -f "$src/$required" ] || { echo "reuse source missing $required: $src" >&2; exit 1; }
|
|
370
|
+
done
|
|
102
371
|
mkdir -p "$(dirname "$dst")"
|
|
103
372
|
cp -R "$src" "$dst"
|
|
104
373
|
echo "[full-pipeline-pair] reused $fid / $arm from $REUSE_CALIBRATED_FROM"
|
|
105
374
|
}
|
|
106
375
|
|
|
376
|
+
if [ -z "$REUSE_CALIBRATED_FROM" ]; then
|
|
377
|
+
mirror_skills
|
|
378
|
+
fi
|
|
379
|
+
|
|
107
380
|
for fid in "${FIXTURES[@]}"; do
|
|
108
381
|
if [ -n "$REUSE_CALIBRATED_FROM" ]; then
|
|
109
382
|
copy_calibrated_arm "$fid" bare
|
|
@@ -129,11 +402,24 @@ headroom_args=(
|
|
|
129
402
|
--run-id "$RUN_ID"
|
|
130
403
|
--bare-max "$BARE_MAX"
|
|
131
404
|
--solo-max "$SOLO_MAX"
|
|
405
|
+
--min-bare-headroom "$MIN_BARE_HEADROOM"
|
|
406
|
+
--min-solo-headroom "$MIN_SOLO_HEADROOM"
|
|
132
407
|
--min-fixtures "$MIN_FIXTURES"
|
|
133
408
|
--out-json "$BENCH_ROOT/results/$RUN_ID/headroom-gate.json"
|
|
134
409
|
--out-md "$BENCH_ROOT/results/$RUN_ID/headroom-gate.md"
|
|
135
410
|
)
|
|
136
|
-
|
|
411
|
+
if ! run_gate_with_report \
|
|
412
|
+
"headroom gate" \
|
|
413
|
+
"$BENCH_ROOT/results/$RUN_ID/headroom-gate.md" \
|
|
414
|
+
python3 "$BENCH_ROOT/scripts/headroom-gate.py" "${headroom_args[@]}"; then
|
|
415
|
+
echo "[full-pipeline-pair] headroom gate failed — pair arm not executed."
|
|
416
|
+
exit 1
|
|
417
|
+
fi
|
|
418
|
+
echo "[full-pipeline-pair] headroom gate passed — executing $PAIR_ARM."
|
|
419
|
+
|
|
420
|
+
if [ -n "$REUSE_CALIBRATED_FROM" ]; then
|
|
421
|
+
mirror_skills
|
|
422
|
+
fi
|
|
137
423
|
|
|
138
424
|
for fid in "${FIXTURES[@]}"; do
|
|
139
425
|
echo "[full-pipeline-pair] ► $fid / $PAIR_ARM"
|
|
@@ -150,13 +436,23 @@ pair_args=(
|
|
|
150
436
|
--run-id "$RUN_ID"
|
|
151
437
|
--bare-max "$BARE_MAX"
|
|
152
438
|
--solo-max "$SOLO_MAX"
|
|
439
|
+
--min-bare-headroom "$MIN_BARE_HEADROOM"
|
|
440
|
+
--min-solo-headroom "$MIN_SOLO_HEADROOM"
|
|
153
441
|
--min-fixtures "$MIN_FIXTURES"
|
|
154
442
|
--min-pair-margin "$MIN_PAIR_MARGIN"
|
|
155
443
|
--pair-arm "$PAIR_ARM"
|
|
444
|
+
--require-hypothesis-trigger
|
|
156
445
|
--out-json "$BENCH_ROOT/results/$RUN_ID/full-pipeline-pair-gate.json"
|
|
157
446
|
--out-md "$BENCH_ROOT/results/$RUN_ID/full-pipeline-pair-gate.md"
|
|
158
447
|
)
|
|
159
448
|
[ -z "$MAX_PAIR_SOLO_WALL_RATIO" ] || pair_args+=(--max-pair-solo-wall-ratio "$MAX_PAIR_SOLO_WALL_RATIO")
|
|
160
449
|
|
|
161
|
-
|
|
162
|
-
|
|
450
|
+
if ! run_gate_with_report \
|
|
451
|
+
"full-pipeline pair gate" \
|
|
452
|
+
"$BENCH_ROOT/results/$RUN_ID/full-pipeline-pair-gate.md" \
|
|
453
|
+
python3 "$BENCH_ROOT/scripts/full-pipeline-pair-gate.py" "${pair_args[@]}"; then
|
|
454
|
+
echo "[full-pipeline-pair] pair gate failed — pair evidence rejected."
|
|
455
|
+
exit 1
|
|
456
|
+
fi
|
|
457
|
+
echo "[full-pipeline-pair] pair gate passed — pair evidence accepted."
|
|
458
|
+
echo "[full-pipeline-pair] release audit: npx devlyn-cli benchmark audit --require-hypothesis-trigger --out-dir /tmp/devlyn-benchmark-audit-strict"
|