Report 013: Release drift, claude-opus-5-5, claude-sonnet-5, claude-sonnet-5-5

Type Release drift report

What moved Claude Sonnet 5.5 on release day, three skills, three runs

Models claude-opus-5-5 claude-sonnet-5 claude-sonnet-5-5

Headline Claude Sonnet 5.5 on the three skills and cases of Reports 009 and 011, beside fresh Claude Sonnet 5 and Claude Opus 5.5 arms on one Claude Code version, each arm run three times.

Receipts 27 linked receipts

Cite this

@techreport{driftproof-013,
  title  = {Report 013: Claude Sonnet 5.5 on release day, three skills, three runs},
  author = {Driftproof},
  year   = {2026},
  type   = {Release drift report},
  note   = {claude-opus-5-5, claude-sonnet-5, claude-sonnet-5-5},
  url    = {https://driftproofhq.com/reports/013/}
}

Report 013: Claude Sonnet 5.5 on release day, three skills, three runs

Release drift report. The first table is the type’s question: the model moves under the skill and the harness stays fixed. The second sets two current models side by side on the same harness, and the third holds the model and moves the harness; the type describes neither, and the limits say how each is read.

Claude Sonnet 5.5 on the three skills and cases of Reports 009 and 011, beside fresh Claude Sonnet 5 and Claude Opus 5.5 arms on one Claude Code version, each arm run three times. Read by the runner’s own comparison of the with-skill arms, run by run: against Claude Sonnet 5, documentation-and-adrs reads separation upward in two of three runs, and code-review-and-quality and git-workflow-and-versioning read separation in zero of six; against Claude Opus 5.5, the three skills read separation in zero of nine.

Every figure below is read from a receipt or a file published beside it, and each block names its files. The target model of the new arm is claude-sonnet-5-5, and every draw of every arm was judged by claude-opus-5. This report measures what a skill adds to each model on these cases. It makes no claim about any model’s coding ability.

Limits, read these first

One case per skill. Each receipt carries a suite of 1, 1, 1 case, in the order code-review-and-quality, git-workflow-and-versioning, documentation-and-adrs, the cases Reports 009 and 011 used. Nothing on this page describes how any of the three skills behaves on any other task, or how any of the three models behaves in general.
Few draws. Each arm drew between 3 and 9 generations, each judged 3 times, and each case’s band is the spread across its draws. With one case there is no spread across cases, and each receipt records its combined uncertainty on the lift as absent, single_case.
Three runs, and a verdict is read in one run at a time. Reports 009 and 011 ran each arm once. This report runs every arm three times, one after another, with the same script and settings, because one run’s figure can move in the next: Report 011’s Claude Opus 5.5 baseline on code-review-and-quality read 0.68, and the same arm in this report’s run 1 read 0.83. Each verdict below is the runner’s comparison of two receipts from one run. Across the three runs the page counts those verdicts; it pools no draws and reads no verdict across runs.
The judge is not one of the three arms, and it is Report 009’s. Every draw was judged by claude-opus-5, as in Reports 009 and 011. It is not the judge the judge policy fixes for Driftproof’s reports, so these figures are comparable with Reports 009 and 011 and not with Reports 001 to 008.
A verdict reads the with-skill arms only. The runner’s comparison of two receipts sets one receipt’s with-skill band against the other’s, per case. The second question is about lift, what the skill adds over the same model without it. The runner does not compare lifts, so the lifts are shown as context and no verdict on this page is read from them.
The harness check moves more than Claude Code. It sets Report 011’s Claude Opus 5.5 receipts, one run on an older Claude Code, beside this report’s three Claude Opus 5.5 runs. The date changed too, and every generation is a fresh draw, so a separation there could not be put down to the Claude Code version alone. Report 011’s sidecars record no per-call effort state, so whether Claude Code applied a per-turn effort there, as it did in all of this report’s Claude Opus 5.5 calls, is not known.
Verification levels. All 27 receipts of this report are TESTED, each answered by a model the surface attested (answered_by.kind model in every one).

What departs from Reports 009 and 011

One Claude Code version for every arm. Every call of every arm ran on Claude Code 2.1.284, installed at the npm integrity the run record states. Report 009 ran on 2.1.272 and Report 011 on 2.1.280. The runner writes no Claude Code version into a receipt, and a receipt is sealed, so each receipt has a sidecar, <receipt>.surface.json, bound to it by its receipt_hash and recording the version each of its calls reported: 2.1.284 in all of them.
Effort and thinking at Claude Code’s defaults. No arm was given an effort or thinking setting, as in Reports 009 and 011. Each call’s first event records whether Claude Code applied a per-turn effort, and the sidecars carry it: claude-sonnet-5-5 true in all 61 of its generation calls across the three runs; claude-sonnet-5 false in all 68 of its generation calls across the three runs; claude-opus-5-5 true in all 75 of its generation calls across the three runs. The judge’s calls read false. So the arms ran on the same surface with each model’s own default, and not at one effort.
The prices are this run’s own. The runner’s registry had no row for claude-sonnet-5-5. This run used a copy of it with that row added at 2 and 10 dollars per million input and output tokens, claude-sonnet-5 at 2 and 10, and claude-opus-5-5 at 4 and 20, each read from the vendor’s pricing page on the day of the run (docs-pricing-snapshot-2026-09-29.json). For claude-sonnet-5 the page’s footnote reads The previously scheduled increase to $3/$15 per million input/output tokens on September 1, 2026 will not occur. The rows are this run’s only; the product’s registry does not carry them. The runner’s registry, config/models.json, prices claude-sonnet-5 at 3 and 15, and each receipt’s pricing_snapshot.source names that file, config/models.json, although its prices are this run’s copy. Prices change the estimated cost below and no score.
A larger call cap across the run. A guard in front of Claude Code allowed 360 calls per run, where Report 011’s allowed 240: nine skill runs where Report 011 ran six. Per skill run the runner’s own cap stands at 80 calls. The guard also stops a run when a call reports any model other than the one it asked for; no run stopped.

What stayed fixed

The inputs are Reports 009 and 011’s. The same SKILL.md bytes (skill.content_hash, the same in all of this report’s receipts and Report 011’s for each skill: code-review-and-quality 13d360d7f786, git-workflow-and-versioning 91c8c72654ee, documentation-and-adrs b67a9f07ed10) and the same suites (suite.suite_hash: code-review-and-quality 5d729f885294, git-workflow-and-versioning 4e74150753d0, documentation-and-adrs 6cce54e7d9d0).

The runner is theirs. Driftproof runner 0.10.1 on the claude-cli surface, with the flags their runs used: at most 80 calls and 8.00 dollars of estimated spend per skill run. It puts the SKILL.md text in the prompt for the with-skill arm and gives the baseline arm the task alone, with no tools in either. It draws generations per arm until the spread settles or a maximum is reached, and the judge scores each draw against the case’s rubric on a continuous 0 to 1 scale.

The judge is theirs. claude-opus-5, with the grading template 82586d1e44f8 in every receipt.

The rule. A case’s band is its mean across draws plus or minus the sample standard deviation across draws: a descriptive spread with no coverage probability. Two with-skill bands separate under the rule when they do not overlap and their means differ by at least the 0.05 effect floor. A case that does not separate is read under spec 035’s rule: when the two arms’ spreads and draws could not have resolved a shift of the floor’s size, it reads not enough draws to conclude at this effect floor, and otherwise no separation detected. Neither is evidence that nothing differs. The verdicts are the runner’s: driftproof diff over each pair, whose outputs are published below.

Did the skills survive the upgrade? Claude Sonnet 5 against Claude Sonnet 5.5

skill	run	Claude Sonnet 5, with skill, mean ± sd across draws	Claude Sonnet 5.5, with skill, mean ± sd across draws	with-skill delta	under the rule	lift (Claude Sonnet 5; Claude Sonnet 5.5), context
`code-review-and-quality`	run 1	0.880 ± 0.006 (3 draws)	0.887 ± 0.025 (3 draws)	+0.007	no separation detected; not enough draws to conclude at this effect floor	+0.302; +0.009
`code-review-and-quality`	run 2	0.881 ± 0.005 (3 draws)	0.900 ± 0.012 (3 draws)	+0.019	no separation detected	+0.207; +0.051
`code-review-and-quality`	run 3	0.657 ± 0.257 (4 draws)	0.896 ± 0.002 (3 draws)	+0.238	no separation detected; not enough draws to conclude at this effect floor	+0.085; +0.034
`code-review-and-quality`	the three runs	0.657 to 0.881	0.887 to 0.900		no separation detected in one of three; no separation detected; not enough draws to conclude at this effect floor in two of three	+0.085 to +0.302; +0.009 to +0.051
`git-workflow-and-versioning`	run 1	0.814 ± 0.028 (3 draws)	0.860 ± 0.009 (3 draws)	+0.046	no separation detected; not enough draws to conclude at this effect floor	+0.012; +0.000
`git-workflow-and-versioning`	run 2	0.851 ± 0.002 (3 draws)	0.863 ± 0.000 (3 draws)	+0.012	no separation detected	+0.020; +0.013
`git-workflow-and-versioning`	run 3	0.831 ± 0.010 (3 draws)	0.863 ± 0.009 (3 draws)	+0.032	no separation detected	-0.004; +0.007
`git-workflow-and-versioning`	the three runs	0.814 to 0.851	0.860 to 0.863		no separation detected in two of three; no separation detected; not enough draws to conclude at this effect floor in one of three	-0.004 to +0.020; +0.000 to +0.013
`documentation-and-adrs`	run 1	0.602 ± 0.004 (3 draws)	0.781 ± 0.141 (6 draws)	+0.178	separation detected, upward	-0.059; +0.208
`documentation-and-adrs`	run 2	0.700 ± 0.098 (5 draws)	0.813 ± 0.094 (7 draws)	+0.113	no separation detected; not enough draws to conclude at this effect floor	-0.006; +0.238
`documentation-and-adrs`	run 3	0.641 ± 0.167 (5 draws)	0.847 ± 0.015 (3 draws)	+0.205	separation detected, upward	-0.085; +0.272
`documentation-and-adrs`	the three runs	0.602 to 0.700	0.781 to 0.847		separation detected, upward in two of three; no separation detected; not enough draws to conclude at this effect floor in one of three	-0.085 to -0.006; +0.208 to +0.272

The with-skill delta is the second with-skill mean minus the first, in the same run. The rule column is the runner’s comparison of the two with-skill bands; the delta and the lifts are context, and no verdict is read from them. A lift is a receipt’s with-skill mean minus its baseline mean. The row for the three runs gives the lowest and highest with-skill mean and lift across the runs, and counts the runner’s readings.

Do the skills help Claude Sonnet 5.5 as much as Claude Opus 5.5?

skill	run	Claude Opus 5.5, with skill, mean ± sd across draws	Claude Sonnet 5.5, with skill, mean ± sd across draws	with-skill delta	under the rule	lift (Claude Opus 5.5; Claude Sonnet 5.5), context
`code-review-and-quality`	run 1	0.907 ± 0.015 (3 draws)	0.887 ± 0.025 (3 draws)	-0.020	no separation detected; not enough draws to conclude at this effect floor	+0.076; +0.009
`code-review-and-quality`	run 2	0.894 ± 0.017 (3 draws)	0.900 ± 0.012 (3 draws)	+0.006	no separation detected; not enough draws to conclude at this effect floor	+0.038; +0.051
`code-review-and-quality`	run 3	0.903 ± 0.015 (3 draws)	0.896 ± 0.002 (3 draws)	-0.008	no separation detected	+0.190; +0.034
`code-review-and-quality`	the three runs	0.894 to 0.907	0.887 to 0.900		no separation detected in one of three; no separation detected; not enough draws to conclude at this effect floor in two of three	+0.038 to +0.190; +0.009 to +0.051
`git-workflow-and-versioning`	run 1	0.861 ± 0.005 (3 draws)	0.860 ± 0.009 (3 draws)	-0.001	no separation detected	+0.353; +0.000
`git-workflow-and-versioning`	run 2	0.857 ± 0.000 (3 draws)	0.863 ± 0.000 (3 draws)	+0.007	no separation detected	+0.371; +0.013
`git-workflow-and-versioning`	run 3	0.856 ± 0.010 (3 draws)	0.863 ± 0.009 (3 draws)	+0.008	no separation detected	+0.276; +0.007
`git-workflow-and-versioning`	the three runs	0.856 to 0.861	0.860 to 0.863		no separation detected in three of three	+0.276 to +0.371; +0.000 to +0.013
`documentation-and-adrs`	run 1	0.650 ± 0.102 (6 draws)	0.781 ± 0.141 (6 draws)	+0.131	no separation detected; not enough draws to conclude at this effect floor	+0.116; +0.208
`documentation-and-adrs`	run 2	0.850 ± 0.007 (3 draws)	0.813 ± 0.094 (7 draws)	-0.037	no separation detected; not enough draws to conclude at this effect floor	+0.300; +0.238
`documentation-and-adrs`	run 3	0.840 ± 0.020 (3 draws)	0.847 ± 0.015 (3 draws)	+0.007	no separation detected; not enough draws to conclude at this effect floor	+0.322; +0.272
`documentation-and-adrs`	the three runs	0.650 to 0.850	0.781 to 0.847		no separation detected; not enough draws to conclude at this effect floor in three of three	+0.116 to +0.322; +0.208 to +0.272

The same columns, with Claude Opus 5.5 first. The question is about lift, and the runner does not compare lifts: the lift column answers it as context, and the rule column reads only the with-skill bands. The vendor’s list prices for the two models are in the section on departures. No vendor statement about these two models is committed with this report, so none is quoted, and nothing on this page tests one.

The harness check: Claude Opus 5.5 in Report 011 against Claude Opus 5.5 in this report

skill	run	Report 011, with skill, mean ± sd across draws	this report, with skill, mean ± sd across draws	with-skill delta	under the rule	lift (Report 011; this report), context
`code-review-and-quality`	run 1	0.910 ± 0.015 (3 draws)	0.907 ± 0.015 (3 draws)	-0.003	no separation detected; not enough draws to conclude at this effect floor	+0.230; +0.076
`code-review-and-quality`	run 2	0.910 ± 0.015 (3 draws)	0.894 ± 0.017 (3 draws)	-0.016	no separation detected; not enough draws to conclude at this effect floor	+0.230; +0.038
`code-review-and-quality`	run 3	0.910 ± 0.015 (3 draws)	0.903 ± 0.015 (3 draws)	-0.007	no separation detected; not enough draws to conclude at this effect floor	+0.230; +0.190
`code-review-and-quality`	the three runs	0.910 (one run)	0.894 to 0.907		no separation detected; not enough draws to conclude at this effect floor in three of three	+0.230; +0.038 to +0.190
`git-workflow-and-versioning`	run 1	0.838 ± 0.027 (3 draws)	0.861 ± 0.005 (3 draws)	+0.023	no separation detected; not enough draws to conclude at this effect floor	+0.538; +0.353
`git-workflow-and-versioning`	run 2	0.838 ± 0.027 (3 draws)	0.857 ± 0.000 (3 draws)	+0.019	no separation detected; not enough draws to conclude at this effect floor	+0.538; +0.371
`git-workflow-and-versioning`	run 3	0.838 ± 0.027 (3 draws)	0.856 ± 0.010 (3 draws)	+0.018	no separation detected; not enough draws to conclude at this effect floor	+0.538; +0.276
`git-workflow-and-versioning`	the three runs	0.838 (one run)	0.856 to 0.861		no separation detected; not enough draws to conclude at this effect floor in three of three	+0.538; +0.276 to +0.371
`documentation-and-adrs`	run 1	0.661 ± 0.148 (6 draws)	0.650 ± 0.102 (6 draws)	-0.011	no separation detected; not enough draws to conclude at this effect floor	+0.094; +0.116
`documentation-and-adrs`	run 2	0.661 ± 0.148 (6 draws)	0.850 ± 0.007 (3 draws)	+0.189	separation detected, upward	+0.094; +0.300
`documentation-and-adrs`	run 3	0.661 ± 0.148 (6 draws)	0.840 ± 0.020 (3 draws)	+0.179	separation detected, upward	+0.094; +0.322
`documentation-and-adrs`	the three runs	0.661 (one run)	0.650 to 0.850		separation detected, upward in two of three; no separation detected; not enough draws to conclude at this effect floor in one of three	+0.094; +0.116 to +0.322

The same columns, with Report 011’s receipt first. Report 011 ran once, so its side of each row is the same receipt in all three runs. Report 011’s receipts are published with that report and linked here, not copied.

Four readings

1. With the skill, Claude Sonnet 5.5 separates upward from Claude Sonnet 5 on documentation-and-adrs in two of three runs, and downward in zero of nine readings. On code-review-and-quality and git-workflow-and-versioning it detects separation in zero of six readings, and in three of those six readings the draws could not have told. What the skill adds is another matter: on code-review-and-quality Claude Sonnet 5’s lift is +0.302, +0.207, +0.085 across the runs and Claude Sonnet 5.5’s +0.009, +0.051, +0.034, because Claude Sonnet 5.5 without the skill already reads 0.878, 0.849, 0.861, against Claude Sonnet 5’s 0.578, 0.674, 0.573. So on these cases the rule reads no downward separation in the upgrade, on any skill in any run, and on code-review-and-quality Claude Sonnet 5.5’s lift is the smaller in every run, from a higher score without it. The runner’s comparison of two receipts reads no lift, so that second part is context, not a verdict.

2. With the skill, the rule separates Claude Sonnet 5.5 from Claude Opus 5.5 in zero of nine readings; the lifts differ where the baselines do. Of the nine readings, five are not enough draws to conclude at this effect floor. On git-workflow-and-versioning Claude Opus 5.5’s lift is +0.353, +0.371, +0.276 and Claude Sonnet 5.5’s +0.000, +0.013, +0.007; with the skill the rule detects separation in zero of three runs, and without it Claude Opus 5.5 reads 0.508, 0.485, 0.579 where Claude Sonnet 5.5 reads 0.860, 0.850, 0.857. On documentation-and-adrs both lifts are above zero in every run: Claude Sonnet 5.5 by +0.208, +0.238, +0.272, Claude Opus 5.5 by +0.116, +0.300, +0.322. So on these cases the skill does not take Claude Sonnet 5.5 to a with-skill score the rule can tell apart from Claude Opus 5.5’s, and how much it adds depends on where each model starts without it. The runner does not compare lifts, and none of this is a statement about either model’s coding ability. Each receipt’s own reading of its lift is on its receipt page; a lift is context here, and none is read as a gain.

3. Claude Opus 5.5 on this report’s Claude Code separates from Report 011’s receipts upward on documentation-and-adrs in some runs and not others. On documentation-and-adrs the rule reads separation upward in two of the three runs, against Report 011’s single run, whose spread there was 0.148. On code-review-and-quality and git-workflow-and-versioning it detects separation in zero of six readings. Report 011 is one run, and this report’s own runs differ from each other, so this is a statement about these draws on two Claude Code versions and two dates, not about Claude Code.

4. The runs differ from each other, which is why there are three. Of the nine skill and comparison pairs, five read a different verdict in one run than in another. Claude Opus 5.5’s baseline on code-review-and-quality read 0.831, 0.857, 0.713 across the three runs. Claude Sonnet 5’s with-skill arm on the same skill read 0.657 ± 0.257 (4 draws) in run 3, against 0.880 ± 0.006 (3 draws) and 0.881 ± 0.005 (3 draws) in the other two. A reading from one run of one case is one reading; the counts above say how often it came back.

Run record

The runs. Each run went arm by arm, claude-sonnet-5-5 first, then claude-sonnet-5, then claude-opus-5-5, one skill at a time, code-review-and-quality first in each. The command file writes a status line after each skill run. The three runs went one after another with the same script and settings. Their log records that the script driving the second and third stopped after the second, in its words three-runs.sh died waiting on the quiet marker, run 3 not started, and that the third then ran, run 3 exit 0. Each receipt’s run.date_utc is written by the runner and is not read here as a start or a finish.

run 1, 20260929T035323Z: nine status lines, each with exit status 0 and one receipt, the last at 2026-09-29T04:37:54Z; 272 calls; 0 unmeasured draws.
run 2, 20260929T043803Z: nine status lines, each with exit status 0 and one receipt, the last at 2026-09-29T05:20:21Z; 272 calls; 0 unmeasured draws.
run 3, 20260929T053800Z: nine status lines, each with exit status 0 and one receipt, the last at 2026-09-29T06:24:35Z; 272 calls; 0 unmeasured draws.

The caps. A guard in front of Claude Code allowed 360 calls per run and 240 seconds per call.

The cost. Metered spend, as the runner reports it, was 0.00 dollars in each skill run’s console output: the claude-cli surface runs on a subscription. The estimated API-equivalent below is every draw’s generation and judge token counts, as each receipt records them, at the prices each receipt froze, with cached input counted at the full input price, as the runner counts it.

run	arm	calls	generation, estimated USD	judge, estimated USD	total, estimated USD
run 1	`claude-sonnet-5-5`	84	0.79	5.59	6.38
run 1	`claude-sonnet-5`	80	0.96	5.31	6.27
run 1	`claude-opus-5-5`	108	2.08	7.24	9.32
run 2	`claude-sonnet-5-5`	88	0.83	5.78	6.62
run 2	`claude-sonnet-5`	88	1.06	5.63	6.70
run 2	`claude-opus-5-5`	96	1.78	6.31	8.09
run 3	`claude-sonnet-5-5`	72	0.68	4.80	5.48
run 3	`claude-sonnet-5`	104	1.28	6.88	8.16
run 3	`claude-opus-5-5`	96	1.88	6.58	8.46
all three runs			11.35	54.14	65.48

Integrity. Each receipt validates with its receipt hash verified, and each sidecar names its receipt’s hash. The raw call records carry the SKILL.md text and stay unpublished; each run’s list of their sha256 is published.

Published evidence

The files this report makes public. For each run, the nine receipts with their summaries, surface sidecars and the runner’s console output, the run record, its registry copy, the status lines and the sha256 of every raw call record; the twenty-seven driftproof diff outputs the verdicts are read from; the log of the three runs; the pricing snapshot; and the call guard. The pricing snapshot records the sha256 of the page it was read from; that page is not published. Each sidecar’s surface_note path is relative to the operator’s run directory; the run record it names is run-<k>/run-record.json here.

docs/reports/013/evidence/diff--harness--run-1--code-review-and-quality.md
sha256 662ac54141c00079b611537b1fc18887dc7f4b6b8e904a1af36c67ec88510476
docs/reports/013/evidence/diff--harness--run-1--documentation-and-adrs.md
sha256 c47cd542a6163b5fe0a0a3555d956bd1b25935fc952b8605371ea25b2f054256
docs/reports/013/evidence/diff--harness--run-1--git-workflow-and-versioning.md
sha256 460b62f91550f49b4a5eb2606dcfcf6c0bd17fc52c99a3e47dcf2bd2b3ac8c3b
docs/reports/013/evidence/diff--harness--run-2--code-review-and-quality.md
sha256 1898c9c7de95fb658c7f24b8bb27b526a7f276db677d9219789313b3c13cdc13
docs/reports/013/evidence/diff--harness--run-2--documentation-and-adrs.md
sha256 5e4aab5e534f0e0baca74c2938f92ebd74c20edef45560c6a5d4a4d0a5a9a358
docs/reports/013/evidence/diff--harness--run-2--git-workflow-and-versioning.md
sha256 a8ec012916f4ea5a22e83ed7985db3e54cbee0675f7d99c0eaed91105722b1de
docs/reports/013/evidence/diff--harness--run-3--code-review-and-quality.md
sha256 c850e21cdb469d9473ee68d3b343030969c3281430d9face8dbedcf75ac38b17
docs/reports/013/evidence/diff--harness--run-3--documentation-and-adrs.md
sha256 18afe91763171920e123c63d1ef2cd35786d9383f8ace6ec33317db9fe0866f9
docs/reports/013/evidence/diff--harness--run-3--git-workflow-and-versioning.md
sha256 2f3a770a3499498622a15daf1f1ae046750ca80f5f56a0d91727f0a75f37d0de
docs/reports/013/evidence/diff--peer--run-1--code-review-and-quality.md
sha256 cdbcb49d16f1a01b9f8105e26a19bfa11a08f932af298ccd0f7e1123c6e31fe4
docs/reports/013/evidence/diff--peer--run-1--documentation-and-adrs.md
sha256 cc2c3e071134716c20a83c3720c6b1495d9684923305c31773936f0135125dbe
docs/reports/013/evidence/diff--peer--run-1--git-workflow-and-versioning.md
sha256 20207e7998faa992a0b611733fbab0db73ea7a806303b7c7f617559ac268da9a
docs/reports/013/evidence/diff--peer--run-2--code-review-and-quality.md
sha256 4969ac4d5de4e36953ffb08f032446d28643c8de51171d66fb91a311149f1369
docs/reports/013/evidence/diff--peer--run-2--documentation-and-adrs.md
sha256 490a3f7c0bc3ff5e22173fe798877cdc366e180ec40c0d2bf2dda3d0e9624098
docs/reports/013/evidence/diff--peer--run-2--git-workflow-and-versioning.md
sha256 4c5e18a1a611c22c7046d51127d58a547caa8eb4aa1d20d1b6cca85b1c038747
docs/reports/013/evidence/diff--peer--run-3--code-review-and-quality.md
sha256 c3944c214613c18b9991d7df0772b0d91b3cea38132533d98bd50790424af748
docs/reports/013/evidence/diff--peer--run-3--documentation-and-adrs.md
sha256 b6ca86734cc00a6d8c6eb02b3da6e776d1fbbc5e09b8a6204cccfbec4be3bd99
docs/reports/013/evidence/diff--peer--run-3--git-workflow-and-versioning.md
sha256 fddc254668aeb6b6b07347bbb65da29d27882e02b74ec3a96029e245a50e68c9
docs/reports/013/evidence/diff--upgrade--run-1--code-review-and-quality.md
sha256 345206d73f871f9bc3db3c41847fd9ee2348410bc01890a7d570f103a26102a0
docs/reports/013/evidence/diff--upgrade--run-1--documentation-and-adrs.md
sha256 e23c97a41a460ea965c3960daea33fbb90accb7d00f022c84e4c1e796295fd22
docs/reports/013/evidence/diff--upgrade--run-1--git-workflow-and-versioning.md
sha256 52ff30649847dbfc2942e88ff8a309b2280062a3f482294c1f19c3f64db3a5ca
docs/reports/013/evidence/diff--upgrade--run-2--code-review-and-quality.md
sha256 80aeeda830067dc3f089f3396b0cc7a73419b6fd52aee7fe21ac96f628c47e5c
docs/reports/013/evidence/diff--upgrade--run-2--documentation-and-adrs.md
sha256 7269c9ea6508c502d51c5845fa0d311f9fdb24f0a28c8a7dd2e483c226ace188
docs/reports/013/evidence/diff--upgrade--run-2--git-workflow-and-versioning.md
sha256 f46ac3eba4725f4e8b35a1abadfd0ab8037adbd8a6d7e1111a905d0f8a9d6b2b
docs/reports/013/evidence/diff--upgrade--run-3--code-review-and-quality.md
sha256 a731f85cf8bb9d9183745b2edc8297438b68a215d0783f9c1381d0a5b884db7e
docs/reports/013/evidence/diff--upgrade--run-3--documentation-and-adrs.md
sha256 3656f5adda6cddd96ff7d28f53f88b4e30bda91ade3c6eb70239496bf3924e40
docs/reports/013/evidence/diff--upgrade--run-3--git-workflow-and-versioning.md
sha256 ef1e8e75f2c2e9bcd78109635a1c54577345f7eb5e8ba0cb876b5335c3046020
docs/reports/013/evidence/docs-pricing-snapshot-2026-09-29.json
sha256 1ce78a21716d7e10a6e27be2b5476e2a7a6541ccf0b26d4036f0c3ed73e6737f
docs/reports/013/evidence/guard.py
sha256 df3922f5b0b09f4c3f1c2d0d1a044b9286367367bf0db6634e8beaaca67f5f1a
docs/reports/013/evidence/run-1/code-review-and-quality-claude-opus-5-5-2026-09-29.json
sha256 08f29e268db4b3d2822d392939e62680b6a5532e1a14d1e78a140db43f41b1fa
docs/reports/013/evidence/run-1/code-review-and-quality-claude-opus-5-5-2026-09-29.stdout.txt
sha256 fcfe4bbd7d55284f702f1c4f441dd506b8babec2116f372cb073c7ef8df5cc3c
docs/reports/013/evidence/run-1/code-review-and-quality-claude-opus-5-5-2026-09-29.summary.md
sha256 552a00d2157f1915ec8108f7a7cddedd6f707d7a9003edadcb9afaaae5148ee4
docs/reports/013/evidence/run-1/code-review-and-quality-claude-opus-5-5-2026-09-29.surface.json
sha256 ef8335899db6163ba635cd246eb5e54e5ec5ebde3306b5d74a688ee1b2a8768c
docs/reports/013/evidence/run-1/code-review-and-quality-claude-sonnet-5-2026-09-29.json
sha256 d86420fa241b6bb11e8fa40a34b2f5811b86a58010ce0d1fac4107ed07cd7137
docs/reports/013/evidence/run-1/code-review-and-quality-claude-sonnet-5-2026-09-29.stdout.txt
sha256 ed9a4b76e1e8fa9b66ad7429e8d08806706be7de119a65dea2b6d094f73df9d2
docs/reports/013/evidence/run-1/code-review-and-quality-claude-sonnet-5-2026-09-29.summary.md
sha256 bfff29c1d7844cabb68d2390916b3d9c944fcc9fe54889b18c8cd28d5cb57437
docs/reports/013/evidence/run-1/code-review-and-quality-claude-sonnet-5-2026-09-29.surface.json
sha256 ec901f2f2be397d1bc29eb94642cdc6380b477e51843ab5db1df25531b31cbd5
docs/reports/013/evidence/run-1/code-review-and-quality-claude-sonnet-5-5-2026-09-29.json
sha256 e56dce9cd680bae50e824234fad7965da3ab1c91384c5e94a5d6dec5ca980ce8
docs/reports/013/evidence/run-1/code-review-and-quality-claude-sonnet-5-5-2026-09-29.stdout.txt
sha256 ddbc0512f261d6db516605923a80f9974a4a98c340fb25e4a19501a47238f496
docs/reports/013/evidence/run-1/code-review-and-quality-claude-sonnet-5-5-2026-09-29.summary.md
sha256 00849c0373b155fc5e2ca681d424e239a515ed5a09930b12292d37cc9ee5ebfb
docs/reports/013/evidence/run-1/code-review-and-quality-claude-sonnet-5-5-2026-09-29.surface.json
sha256 bd6cfb92a4cf20d079fa487c3906511b2e66d4103784ddc4be6aa97de2f8460b
docs/reports/013/evidence/run-1/documentation-and-adrs-claude-opus-5-5-2026-09-29.json
sha256 b4b0edd02ebf02b2628d5d71397ef59e4b6d06eb1117a93acef70fb902603fce
docs/reports/013/evidence/run-1/documentation-and-adrs-claude-opus-5-5-2026-09-29.stdout.txt
sha256 178bbe4bf19f10bba2efa5641e2dbae6c41b00781fc53bbe68f7779cd6542359
docs/reports/013/evidence/run-1/documentation-and-adrs-claude-opus-5-5-2026-09-29.summary.md
sha256 86c6db0601582364081e162453e28ef5ae370cc880af7d9094d24fe8e2b5884d
docs/reports/013/evidence/run-1/documentation-and-adrs-claude-opus-5-5-2026-09-29.surface.json
sha256 694d2b82ab5a403d32d1e76fe3b382e2e5ad75d08523438518d5b3547ad81cda
docs/reports/013/evidence/run-1/documentation-and-adrs-claude-sonnet-5-2026-09-29.json
sha256 6b208bf92c02c15b9893c22268e3b6e293c42d483c528d76f6abf12b837e0df9
docs/reports/013/evidence/run-1/documentation-and-adrs-claude-sonnet-5-2026-09-29.stdout.txt
sha256 2d6ba28f38d976cc5ffbf85efe609bd1b7ebfdc93979ed44df4e31ae454976b2
docs/reports/013/evidence/run-1/documentation-and-adrs-claude-sonnet-5-2026-09-29.summary.md
sha256 49ffb46b251a8c0234453bb0ac1bf7eb409255fff78d8c80320e73b9ff511acb
docs/reports/013/evidence/run-1/documentation-and-adrs-claude-sonnet-5-2026-09-29.surface.json
sha256 a07e925a39c45dfb402b34e4595ccf8740d1e10f8a861669113328528ee92a0c
docs/reports/013/evidence/run-1/documentation-and-adrs-claude-sonnet-5-5-2026-09-29.json
sha256 c4a435399ea32d2f762a2c27ba3d0046275127a331fdc632aeea10693d76dba1
docs/reports/013/evidence/run-1/documentation-and-adrs-claude-sonnet-5-5-2026-09-29.stdout.txt
sha256 e82b2e390c78a6f74fdad3a672efbba1df7b78a801142c0e226b706463525124
docs/reports/013/evidence/run-1/documentation-and-adrs-claude-sonnet-5-5-2026-09-29.summary.md
sha256 662939c4c41c972e6f5900ad03e15d59de9432e9fd716dbafe45b1eaefa8e0ef
docs/reports/013/evidence/run-1/documentation-and-adrs-claude-sonnet-5-5-2026-09-29.surface.json
sha256 71efa11ed6f30cd09e76f94ebcc6064dc0829251ea883c0c9f948b3e5839535c
docs/reports/013/evidence/run-1/driftproof-calls.SHA256SUMS
sha256 32128aaf617a1bc32346281dbbbec5da30839df2d414bf856168f1e6bc695ee3
docs/reports/013/evidence/run-1/git-workflow-and-versioning-claude-opus-5-5-2026-09-29.json
sha256 c91fe2fc265205bdb55d428d056d2b1e873d25a60c99e71264c792333f5bfd5c
docs/reports/013/evidence/run-1/git-workflow-and-versioning-claude-opus-5-5-2026-09-29.stdout.txt
sha256 ac21ced2e9ba51224825e4958c4a66b538dcb5e928b3ebe4727a30e8c694ce28
docs/reports/013/evidence/run-1/git-workflow-and-versioning-claude-opus-5-5-2026-09-29.summary.md
sha256 2e038be3d4356c0b3d30c184236ace09a1a86a985a6d788a4ebcf9de80ac0c83
docs/reports/013/evidence/run-1/git-workflow-and-versioning-claude-opus-5-5-2026-09-29.surface.json
sha256 9afdfcde3911e4a968924dca2b99f40ba8dbaf3731326485a6d972ac8a5c7e7c
docs/reports/013/evidence/run-1/git-workflow-and-versioning-claude-sonnet-5-2026-09-29.json
sha256 da3989e0a7943ddac208ab7e2d938d807ca4679d8a304787db2fab7315b4af42
docs/reports/013/evidence/run-1/git-workflow-and-versioning-claude-sonnet-5-2026-09-29.stdout.txt
sha256 dba298bf69105219553a2c5f7c11aebbbb579d0272c48cf01437dc0f2d98b37b
docs/reports/013/evidence/run-1/git-workflow-and-versioning-claude-sonnet-5-2026-09-29.summary.md
sha256 1c1bc5007f995a53fb7072a27aff008dd85ec8e0a322276e12028230bc09fe97
docs/reports/013/evidence/run-1/git-workflow-and-versioning-claude-sonnet-5-2026-09-29.surface.json
sha256 504b2dbbf2243f820a70e9a85cee108d2df717be074ddb72c2fde1eb83919c38
docs/reports/013/evidence/run-1/git-workflow-and-versioning-claude-sonnet-5-5-2026-09-29.json
sha256 4c47ba8f4b8ac14bdc20ec79d98c3d66f58d776f80f32d87ffc4cdee7f409d44
docs/reports/013/evidence/run-1/git-workflow-and-versioning-claude-sonnet-5-5-2026-09-29.stdout.txt
sha256 2752b94a92d4e859662373a3289528d87a717ca8487be4e38b5460fcbae1183e
docs/reports/013/evidence/run-1/git-workflow-and-versioning-claude-sonnet-5-5-2026-09-29.summary.md
sha256 e6a29f74445d90cb4f809696a66df6b96d3bf2020a3b0d22a098d806bb10f304
docs/reports/013/evidence/run-1/git-workflow-and-versioning-claude-sonnet-5-5-2026-09-29.surface.json
sha256 adf7f7793690f0bed9152062f4a7638beb6e9f3779197d32c6779f76950bb02b
docs/reports/013/evidence/run-1/registry.json
sha256 0c08c778c7bdd4b9ab02fdf6c4cc8c91686eb315edff83486a1ca7bc0b391598
docs/reports/013/evidence/run-1/run-record.json
sha256 fe50e000dc186ebe990dff370a33ec12f31dcb3967579db34f2c9b9af5358609
docs/reports/013/evidence/run-1/status.jsonl
sha256 a395f40c2feb0168458f1134217eabd8e21f243b5c23c6d6c193120d8a01fc6b
docs/reports/013/evidence/run-2/code-review-and-quality-claude-opus-5-5-2026-09-29.json
sha256 b392e85afdb35112c4a8c279621f893b443116d651b62b5f0157fa08e5f6f6c3
docs/reports/013/evidence/run-2/code-review-and-quality-claude-opus-5-5-2026-09-29.stdout.txt
sha256 f380818a45261d9335578ef03cc63210cb7717891bf13a56be7bcb4252d9fbc8
docs/reports/013/evidence/run-2/code-review-and-quality-claude-opus-5-5-2026-09-29.summary.md
sha256 213a521c9403193e08ee7feed6ffc8887b8c448fbf3366be5622560abc70903c
docs/reports/013/evidence/run-2/code-review-and-quality-claude-opus-5-5-2026-09-29.surface.json
sha256 dedb3462505a1497144aec0687bba55e51fe6cc12df4d0e32a054d8dfcd91a28
docs/reports/013/evidence/run-2/code-review-and-quality-claude-sonnet-5-2026-09-29.json
sha256 884cc2914c0c084c55bbf0b697c92cf7de4f266d3f0383e20bc760a1dea2a5cf
docs/reports/013/evidence/run-2/code-review-and-quality-claude-sonnet-5-2026-09-29.stdout.txt
sha256 38a443da83fd11413c1bb6206d431071993e0e64429aee07a925e13210df6c1a
docs/reports/013/evidence/run-2/code-review-and-quality-claude-sonnet-5-2026-09-29.summary.md
sha256 761053f48746e67b09b32d0a9c72a1f83e64912248a6745e8efe14f4beda81f6
docs/reports/013/evidence/run-2/code-review-and-quality-claude-sonnet-5-2026-09-29.surface.json
sha256 ec3000d9534b3d1317b381eaa6ca915d8fcd811ed03762685b168cd09e5cb6e3
docs/reports/013/evidence/run-2/code-review-and-quality-claude-sonnet-5-5-2026-09-29.json
sha256 9591e1083c6b2bf0a01d2301798ef297983930bb8931c6db06cb7b64c533e36b
docs/reports/013/evidence/run-2/code-review-and-quality-claude-sonnet-5-5-2026-09-29.stdout.txt
sha256 b5b5d58865ac7e6fa4bcd4fe009dfd98c2a08de895e5ad7ef3acad81f6d41eb3
docs/reports/013/evidence/run-2/code-review-and-quality-claude-sonnet-5-5-2026-09-29.summary.md
sha256 1907f26ef0feddec5428c85f43999f576e11deff4d92dc6748d9c5c570d8a4dd
docs/reports/013/evidence/run-2/code-review-and-quality-claude-sonnet-5-5-2026-09-29.surface.json
sha256 e67ed4dddb76f987349f6bf93121360068398d260e85a77297e43afe45aa98b7
docs/reports/013/evidence/run-2/documentation-and-adrs-claude-opus-5-5-2026-09-29.json
sha256 8e6215d0109a6cbb77fc4ef4942b05b68ce7116c042f35eea57b5bc6c5f32991
docs/reports/013/evidence/run-2/documentation-and-adrs-claude-opus-5-5-2026-09-29.stdout.txt
sha256 3de5eeb8add92621b6bcf853c27685ed897aca7f41b61c6819a7a4bc456c3a50
docs/reports/013/evidence/run-2/documentation-and-adrs-claude-opus-5-5-2026-09-29.summary.md
sha256 dc9aa061efbd3be7a1999e4238c730eca7643c1304d559210ab46d9f214469cd
docs/reports/013/evidence/run-2/documentation-and-adrs-claude-opus-5-5-2026-09-29.surface.json
sha256 40507d639cacdf9f2ad69cc2948114ab58ea27d70ef9cc43b6557c9926b5d2f6
docs/reports/013/evidence/run-2/documentation-and-adrs-claude-sonnet-5-2026-09-29.json
sha256 ae3253a7e5ab05e7a8be9175699c81de8d4173b7f8470c7635a19fe8387d49be
docs/reports/013/evidence/run-2/documentation-and-adrs-claude-sonnet-5-2026-09-29.stdout.txt
sha256 f486a98be5fd78c7ea8cfc448bc577723d661a9ba002bb7fa2c82536f943e3c5
docs/reports/013/evidence/run-2/documentation-and-adrs-claude-sonnet-5-2026-09-29.summary.md
sha256 6b614abf92eee3331157c092f24eed481c5f419f314b94b6c81f2faf3564962d
docs/reports/013/evidence/run-2/documentation-and-adrs-claude-sonnet-5-2026-09-29.surface.json
sha256 e9363159a13af626508340cff4592ca242ad635c97e66e677fab4772a4178723
docs/reports/013/evidence/run-2/documentation-and-adrs-claude-sonnet-5-5-2026-09-29.json
sha256 da6f12f6ae042b7b03c9aeb5514c0d4ca5d9bf764f4cc22acef282987dd3d8a8
docs/reports/013/evidence/run-2/documentation-and-adrs-claude-sonnet-5-5-2026-09-29.stdout.txt
sha256 290715a763095052b0ae15a8ac1934475e613ad49f82f796d2753a7198176aba
docs/reports/013/evidence/run-2/documentation-and-adrs-claude-sonnet-5-5-2026-09-29.summary.md
sha256 bc8baf585c6982088bf412291687991656d557555abf07ae60ee1a895d43bc0d
docs/reports/013/evidence/run-2/documentation-and-adrs-claude-sonnet-5-5-2026-09-29.surface.json
sha256 e3ddcdc17f20211c52681d8e45cc2279dbb62c23d2c8a8ffc9e5a6cee821cd6f
docs/reports/013/evidence/run-2/driftproof-calls.SHA256SUMS
sha256 a8dd3ad53f31beff0c555881fdf132be542dfda4491726a1ca0834aae610f16a
docs/reports/013/evidence/run-2/git-workflow-and-versioning-claude-opus-5-5-2026-09-29.json
sha256 524b8a95e8bf93452af0f6614f203d4f9b70d36bb6eceb6882e152a25951bf97
docs/reports/013/evidence/run-2/git-workflow-and-versioning-claude-opus-5-5-2026-09-29.stdout.txt
sha256 4f70fe0ee115e99926a88df381a34724c5fdf05f54ef2ad92c9d9cc64ab3ca0e
docs/reports/013/evidence/run-2/git-workflow-and-versioning-claude-opus-5-5-2026-09-29.summary.md
sha256 d78bb1a92580420de7a1b6c84215adc5a0ae8fed9305569dbba9a482dcdcd467
docs/reports/013/evidence/run-2/git-workflow-and-versioning-claude-opus-5-5-2026-09-29.surface.json
sha256 1d5b20091d869eebb0bd8ecc24b88079562ee081624d9bc1ced692366dfe5e06
docs/reports/013/evidence/run-2/git-workflow-and-versioning-claude-sonnet-5-2026-09-29.json
sha256 7781543798f3ba27589cf631531924ab6515fff71d6e57f50d8bef5b21187717
docs/reports/013/evidence/run-2/git-workflow-and-versioning-claude-sonnet-5-2026-09-29.stdout.txt
sha256 a6bac7323a2798d202ffee4c6cda9bd84a4f31ff49f9cc0b311e56c41c81b294
docs/reports/013/evidence/run-2/git-workflow-and-versioning-claude-sonnet-5-2026-09-29.summary.md
sha256 7a949e24607b22439068221fa0164acd0cc108af503e79a3a78d04f1a0c719c9
docs/reports/013/evidence/run-2/git-workflow-and-versioning-claude-sonnet-5-2026-09-29.surface.json
sha256 c465c6e9b4a15c32e4755ee342016619456ceef37a3ff26f9da60feaeee1d37a
docs/reports/013/evidence/run-2/git-workflow-and-versioning-claude-sonnet-5-5-2026-09-29.json
sha256 2fbc0b99fb4c60ac4c456882a225f27b492d7007b9fce06abad295e6250c9a5e
docs/reports/013/evidence/run-2/git-workflow-and-versioning-claude-sonnet-5-5-2026-09-29.stdout.txt
sha256 44edbe334b0d7f8ca69443106a5835c7aab30116d79d359942f11e24604f5cca
docs/reports/013/evidence/run-2/git-workflow-and-versioning-claude-sonnet-5-5-2026-09-29.summary.md
sha256 afac75e0eb905cf6db3d57446b8d597da3160a46d64f9267da9ab87e223e2d40
docs/reports/013/evidence/run-2/git-workflow-and-versioning-claude-sonnet-5-5-2026-09-29.surface.json
sha256 119c6574a827f9a03087bc3794041cb556260484092476f98eef3616260dc965
docs/reports/013/evidence/run-2/registry.json
sha256 0c08c778c7bdd4b9ab02fdf6c4cc8c91686eb315edff83486a1ca7bc0b391598
docs/reports/013/evidence/run-2/run-record.json
sha256 ab75be120213302e08e3d37dcef33d499becb6043438569b45110651e6dcfc00
docs/reports/013/evidence/run-2/status.jsonl
sha256 5f21873c496cbbf64b77b1941b986df0524bed109289f255e710604f9c129668
docs/reports/013/evidence/run-3/code-review-and-quality-claude-opus-5-5-2026-09-29.json
sha256 20f9e75da5cf6ee67d44829c12833bf22b28fc846d3d9530179ff30946818ad9
docs/reports/013/evidence/run-3/code-review-and-quality-claude-opus-5-5-2026-09-29.stdout.txt
sha256 b896333abd8746713d624d91af2921b45bfda862e6fe27d59ec5e0b62f125784
docs/reports/013/evidence/run-3/code-review-and-quality-claude-opus-5-5-2026-09-29.summary.md
sha256 11d3575f5db4a7b73aba25fdbb7564fb386b6caa92ee21f0d42d7235e6e2068a
docs/reports/013/evidence/run-3/code-review-and-quality-claude-opus-5-5-2026-09-29.surface.json
sha256 10ba265855606481dd870e13c54886c63d34d0bb4608022e85f1b41b5c31fdbd
docs/reports/013/evidence/run-3/code-review-and-quality-claude-sonnet-5-2026-09-29.json
sha256 765dcd44027dcabfb494fa3aed02fcb845651c048ff8eb2f24ffe54c7e100643
docs/reports/013/evidence/run-3/code-review-and-quality-claude-sonnet-5-2026-09-29.stdout.txt
sha256 ea48c870768e953056c763b75026019a2830d74d1d4d02d67cdae4eec6874c16
docs/reports/013/evidence/run-3/code-review-and-quality-claude-sonnet-5-2026-09-29.summary.md
sha256 c1b20621b5ba8e2f7b2ecc63627cf62b05efba6606fac6ffe7a52810d12addb8
docs/reports/013/evidence/run-3/code-review-and-quality-claude-sonnet-5-2026-09-29.surface.json
sha256 a068ffc49f4b4d4a4237abc722982bdca76e4bc6957ad6d2ce67581c11bfbce5
docs/reports/013/evidence/run-3/code-review-and-quality-claude-sonnet-5-5-2026-09-29.json
sha256 4bdba1d1af2d8e2e5cdf713674ad24d2203d833bd43ce9dfec96ecdf12a0ab77
docs/reports/013/evidence/run-3/code-review-and-quality-claude-sonnet-5-5-2026-09-29.stdout.txt
sha256 b5bc66734d8dbb38021db1fb5b8f5de3da76415a3210ac291d16d0e63a7a4141
docs/reports/013/evidence/run-3/code-review-and-quality-claude-sonnet-5-5-2026-09-29.summary.md
sha256 6c850957a1b51c618e1801207a40c401316681b553f7a76d45c5c3819e211500
docs/reports/013/evidence/run-3/code-review-and-quality-claude-sonnet-5-5-2026-09-29.surface.json
sha256 2d3bcda4c7d1c9d974793aef08fe9212b26a9874ad8e902a8f982549f26ba5bb
docs/reports/013/evidence/run-3/documentation-and-adrs-claude-opus-5-5-2026-09-29.json
sha256 e3c92a942bdfc8aec994cd20394a514bb2284bf5e72b6a15da74f2f711cd3c8f
docs/reports/013/evidence/run-3/documentation-and-adrs-claude-opus-5-5-2026-09-29.stdout.txt
sha256 83b99eaa0819c707ec3ac66ed2409ae54bfc2ab0612da017d93161da47a48857
docs/reports/013/evidence/run-3/documentation-and-adrs-claude-opus-5-5-2026-09-29.summary.md
sha256 2a98efaf6281b42159af347c7d90a2343d900a9fa88363fede2552c1b3a63220
docs/reports/013/evidence/run-3/documentation-and-adrs-claude-opus-5-5-2026-09-29.surface.json
sha256 75f5a859feeb09859952d67b32e91a9373e9dc9dca917e0ac28adfc1165f31fb
docs/reports/013/evidence/run-3/documentation-and-adrs-claude-sonnet-5-2026-09-29.json
sha256 d12e0d3da243370c5e452a05caa7c032d9ba4ddaf1cb4dcd31dfbaf2d9d65a46
docs/reports/013/evidence/run-3/documentation-and-adrs-claude-sonnet-5-2026-09-29.stdout.txt
sha256 6bb80dcd685b098a7a6e8c597aa7ba6bd91ddf57237575a8b9fe70dd3687f99a
docs/reports/013/evidence/run-3/documentation-and-adrs-claude-sonnet-5-2026-09-29.summary.md
sha256 a8094cbde3faeda2545e418efaf88ceddce2b16e93c2d6aae2b9df7526cfdb17
docs/reports/013/evidence/run-3/documentation-and-adrs-claude-sonnet-5-2026-09-29.surface.json
sha256 3af7094f24c36ecdc13a9a1f368b3ae6d3d4ca222e3c5b5780105ea1e34175d3
docs/reports/013/evidence/run-3/documentation-and-adrs-claude-sonnet-5-5-2026-09-29.json
sha256 f6ddaf3d2791aa3047e47a7da1eb451cfc84903a7f63dff70d8c8841fa49496f
docs/reports/013/evidence/run-3/documentation-and-adrs-claude-sonnet-5-5-2026-09-29.stdout.txt
sha256 fe3d904e700df23ac5ffb602e2d2f971d3dd8bed01d3376b3196283b2b40979c
docs/reports/013/evidence/run-3/documentation-and-adrs-claude-sonnet-5-5-2026-09-29.summary.md
sha256 8330839d521f72008936893acdcd95707b269c39f74b8a7f3a4ccb8cc445a8c9
docs/reports/013/evidence/run-3/documentation-and-adrs-claude-sonnet-5-5-2026-09-29.surface.json
sha256 8bc78743e82309b2f1ddd746dce80c0b2f4be4dd7c31442e5c6312885d0e7a25
docs/reports/013/evidence/run-3/driftproof-calls.SHA256SUMS
sha256 4025b9b2c1ca7bb8cf7d9513302e1834ad4d69933b9c699efb92372e9883391b
docs/reports/013/evidence/run-3/git-workflow-and-versioning-claude-opus-5-5-2026-09-29.json
sha256 251ec0adbfc58ed88b44c33c8740f44c1f5a6b7233d7542c773f23a8b97624ff
docs/reports/013/evidence/run-3/git-workflow-and-versioning-claude-opus-5-5-2026-09-29.stdout.txt
sha256 a38055db814bf87cef247fb13be2e4c9af308c1d02dd2b2f6786229829a726c1
docs/reports/013/evidence/run-3/git-workflow-and-versioning-claude-opus-5-5-2026-09-29.summary.md
sha256 253146a035f37d8be7aec66b883938d20a9a0154365fad3ec6d5da8ccad040aa
docs/reports/013/evidence/run-3/git-workflow-and-versioning-claude-opus-5-5-2026-09-29.surface.json
sha256 0aa24b6476f21099009c9e629d4a62a3eeb5589d61e11ed2204cbcb603386e7e
docs/reports/013/evidence/run-3/git-workflow-and-versioning-claude-sonnet-5-2026-09-29.json
sha256 05b8de68aaf45e5dffc2cee2f0ace706b1a2a189d6e5a0a1d1116177db666f3e
docs/reports/013/evidence/run-3/git-workflow-and-versioning-claude-sonnet-5-2026-09-29.stdout.txt
sha256 02bb7c591ab182d1009e023b05b763751498ec4de648b4f19e676a623c2c3dc9
docs/reports/013/evidence/run-3/git-workflow-and-versioning-claude-sonnet-5-2026-09-29.summary.md
sha256 1677da13b885825e9926f593805499ed02f4eedaf5eb9fda22fef5834c9d612b
docs/reports/013/evidence/run-3/git-workflow-and-versioning-claude-sonnet-5-2026-09-29.surface.json
sha256 de8e87d7055c4544cafee246c01272b67d5915ab5a7debcf0cd2d41c27c23d02
docs/reports/013/evidence/run-3/git-workflow-and-versioning-claude-sonnet-5-5-2026-09-29.json
sha256 9a327d6c7adc1925919ca25ef284477b61abf2711ac0d8d6443b4e219e397256
docs/reports/013/evidence/run-3/git-workflow-and-versioning-claude-sonnet-5-5-2026-09-29.stdout.txt
sha256 4121bf0a7e5eb31e2004ca8def499878e80603064bef1cdba22546c757dc3fdf
docs/reports/013/evidence/run-3/git-workflow-and-versioning-claude-sonnet-5-5-2026-09-29.summary.md
sha256 11012191bb087ffb7ec5e28ecb601ebae67c163ddb4ad6fb3e8386b893c7bbb8
docs/reports/013/evidence/run-3/git-workflow-and-versioning-claude-sonnet-5-5-2026-09-29.surface.json
sha256 c14a01e2cf76f415fb9fcbef614075a179c8420ea496e742044e972c1972350a
docs/reports/013/evidence/run-3/registry.json
sha256 0c08c778c7bdd4b9ab02fdf6c4cc8c91686eb315edff83486a1ca7bc0b391598
docs/reports/013/evidence/run-3/run-record.json
sha256 1b3038538b5f875ba99a973752b66b0f00116aefb14db2e10e82ffb9a33a8576
docs/reports/013/evidence/run-3/status.jsonl
sha256 5c3d21c3f8b99acf489023ae58da801b54abead3b8163f69d2bbd548d1abc530
docs/reports/013/evidence/runs-log.txt
sha256 74b4c6c54f010b5287bdaf0d8be4ea541321d521f176a7dbdb54cbd82dbd66f1

Report 011’s three Claude Opus 5.5 receipts, the other side of the harness check, are published with that report: code-review-and-quality, git-workflow-and-versioning, documentation-and-adrs. Validate a receipt with npx driftproof validate <file>. Each copy here is byte-identical to its twin under specs/121-report-013-sonnet-5-5/, and each sidecar names the receipt it belongs to and that receipt’s hash.

Receipts

Every receipt this report rests on, each one resolvable. Nine for each run, one per arm and skill, the same bytes as the evidence copies above.

code-review-and-quality@claude-sonnet-5-5 (run 1): receipts/report-013/run-1/code-review-and-quality-claude-sonnet-5-5-2026-09-29.json
git-workflow-and-versioning@claude-sonnet-5-5 (run 1): receipts/report-013/run-1/git-workflow-and-versioning-claude-sonnet-5-5-2026-09-29.json
documentation-and-adrs@claude-sonnet-5-5 (run 1): receipts/report-013/run-1/documentation-and-adrs-claude-sonnet-5-5-2026-09-29.json
code-review-and-quality@claude-sonnet-5 (run 1): receipts/report-013/run-1/code-review-and-quality-claude-sonnet-5-2026-09-29.json
git-workflow-and-versioning@claude-sonnet-5 (run 1): receipts/report-013/run-1/git-workflow-and-versioning-claude-sonnet-5-2026-09-29.json
documentation-and-adrs@claude-sonnet-5 (run 1): receipts/report-013/run-1/documentation-and-adrs-claude-sonnet-5-2026-09-29.json
code-review-and-quality@claude-opus-5-5 (run 1): receipts/report-013/run-1/code-review-and-quality-claude-opus-5-5-2026-09-29.json
git-workflow-and-versioning@claude-opus-5-5 (run 1): receipts/report-013/run-1/git-workflow-and-versioning-claude-opus-5-5-2026-09-29.json
documentation-and-adrs@claude-opus-5-5 (run 1): receipts/report-013/run-1/documentation-and-adrs-claude-opus-5-5-2026-09-29.json
code-review-and-quality@claude-sonnet-5-5 (run 2): receipts/report-013/run-2/code-review-and-quality-claude-sonnet-5-5-2026-09-29.json
git-workflow-and-versioning@claude-sonnet-5-5 (run 2): receipts/report-013/run-2/git-workflow-and-versioning-claude-sonnet-5-5-2026-09-29.json
documentation-and-adrs@claude-sonnet-5-5 (run 2): receipts/report-013/run-2/documentation-and-adrs-claude-sonnet-5-5-2026-09-29.json
code-review-and-quality@claude-sonnet-5 (run 2): receipts/report-013/run-2/code-review-and-quality-claude-sonnet-5-2026-09-29.json
git-workflow-and-versioning@claude-sonnet-5 (run 2): receipts/report-013/run-2/git-workflow-and-versioning-claude-sonnet-5-2026-09-29.json
documentation-and-adrs@claude-sonnet-5 (run 2): receipts/report-013/run-2/documentation-and-adrs-claude-sonnet-5-2026-09-29.json
code-review-and-quality@claude-opus-5-5 (run 2): receipts/report-013/run-2/code-review-and-quality-claude-opus-5-5-2026-09-29.json
git-workflow-and-versioning@claude-opus-5-5 (run 2): receipts/report-013/run-2/git-workflow-and-versioning-claude-opus-5-5-2026-09-29.json
documentation-and-adrs@claude-opus-5-5 (run 2): receipts/report-013/run-2/documentation-and-adrs-claude-opus-5-5-2026-09-29.json
code-review-and-quality@claude-sonnet-5-5 (run 3): receipts/report-013/run-3/code-review-and-quality-claude-sonnet-5-5-2026-09-29.json
git-workflow-and-versioning@claude-sonnet-5-5 (run 3): receipts/report-013/run-3/git-workflow-and-versioning-claude-sonnet-5-5-2026-09-29.json
documentation-and-adrs@claude-sonnet-5-5 (run 3): receipts/report-013/run-3/documentation-and-adrs-claude-sonnet-5-5-2026-09-29.json
code-review-and-quality@claude-sonnet-5 (run 3): receipts/report-013/run-3/code-review-and-quality-claude-sonnet-5-2026-09-29.json
git-workflow-and-versioning@claude-sonnet-5 (run 3): receipts/report-013/run-3/git-workflow-and-versioning-claude-sonnet-5-2026-09-29.json
documentation-and-adrs@claude-sonnet-5 (run 3): receipts/report-013/run-3/documentation-and-adrs-claude-sonnet-5-2026-09-29.json
code-review-and-quality@claude-opus-5-5 (run 3): receipts/report-013/run-3/code-review-and-quality-claude-opus-5-5-2026-09-29.json
git-workflow-and-versioning@claude-opus-5-5 (run 3): receipts/report-013/run-3/git-workflow-and-versioning-claude-opus-5-5-2026-09-29.json
documentation-and-adrs@claude-opus-5-5 (run 3): receipts/report-013/run-3/documentation-and-adrs-claude-opus-5-5-2026-09-29.json

Validate any of them with npx driftproof validate <file>, and reproduce a verdict with npx driftproof diff <first> <second> over two receipts of one run.