Rendered from docs/probe-results/additive-screening-of-2026-10-18-scenario-3.md in the Headwater
corpus. Every document on this half of the site is typed by the taxonomy
the descriptor names: corpus.json.
The result of docs/probe-runs/additive-screening-of-2026-10-18-scenario-3.md
A probe result is a function of committed inputs and of nothing else: the transcript at docs/probe-runs/additive-screening-of-2026-10-18-scenario-3.md and the state it stands at, the expectations the probes it names declare, the version of the grader that evaluated them, and for a compared arm the result of the other arm. Fetch those and this file comes back.
The run this transcript recorded
A additive run in the scenario-3 arm, on claude-sonnet-5 at 2026-10-18. served version claude-sonnet-5, tree sha256:7d84079327a6496ad99d1047ec2c9d0f82a810c3765b0ae22a4ae3ea97be60e8, selection sha256:0e2113cd81af83407d78a1964326c730f04457f9a81c929f9e32bfc73585bda3, read set sha256:30284342a7371374ee5ded1720314683046f3f9f7b372ecbb70ff7e6f3e8b5d3, seed 0, harness 0.6.0. realized cost $22.32, which the adaptive layer reads as the cost of its own instrument.
72 events over 4 probes, in 72 sessions and 790 tool calls.
The engine confirmed the taxonomy, that every member of the run identity is present, the membership of every probe named, that no key outside the closed set appears, and that a realized cost was recorded. Present is not confirmed. This page names the lock, the selection and the read set the transcript recorded, and it compares none of them with the tree in front of a reader. headwater generate and headwater probe stale name what moved since the recording, and neither one fails for it. It graded nothing: a verdict is a function of this transcript, the expectations these probes declare and a grader version, and headwater probe grade is the verb that holds all three.
The selection this transcript names is sha256:0e2113cd81af83407d78a1964326c730f04457f9a81c929f9e32bfc73585bda3, and the probes graded below are the probes it was planned over, as this corpus declares them now. A probe added to this corpus after the recording is not a session this run owed. The tree, the seed and the harness above are provenance and nothing compares them.
The read_set digest above covers every probe of the selection and every document one of them examines, by path and content. This file names it as provenance and compares it with nothing, and the same holds for the lock and the selection: an edit to a document, a new probe or a moved lock leaves these bytes alone. headwater generate names each result whose recorded lock, read set or selection moved, and headwater probe stale takes the digest and reports which recorded results a change voided. Neither one fails for it.
Graded by grader 0.5.1. A additive run in the scenario-3 arm, on claude-sonnet-5 at served version claude-sonnet-5.
The verdicts
- HW-PROBE-a-cold-agent-reaches-the-governing-document-through-the-corpus-descriptor (discovery, expects opened)
session L4-additive-scenario-3-p1-r1: satisfied — event 1, call 18:
Grep /mnt/build/hw-1775/screening/a/ws/L4-additive-scenario-3-p1-r1/docs/spec/12-check-layer.md; no call to a tool of an MCP server session L4-additive-scenario-3-p1-r10: not satisfied — 18 recorded calls, and none named any of the 1 document; no call to a tool of an MCP server session L4-additive-scenario-3-p1-r11: satisfied — event 3, call 4:Read /mnt/build/hw-1775/screening/a/ws/L4-additive-scenario-3-p1-r11/docs/spec/12-check-layer.md; no call to a tool of an MCP server session L4-additive-scenario-3-p1-r12: not satisfied — 16 recorded calls, and none named any of the 1 document; no call to a tool of an MCP server session L4-additive-scenario-3-p1-r13: not satisfied — 23 recorded calls, and none named any of the 1 document; no call to a tool of an MCP server session L4-additive-scenario-3-p1-r14: not satisfied — 15 recorded calls, and none named any of the 1 document; no call to a tool of an MCP server session L4-additive-scenario-3-p1-r15: satisfied — event 7, call 4:Read /mnt/build/hw-1775/screening/a/ws/L4-additive-scenario-3-p1-r15/docs/spec/12-check-layer.md; no call to a tool of an MCP server session L4-additive-scenario-3-p1-r16: not satisfied — 15 recorded calls, and none named any of the 1 document; no call to a tool of an MCP server session L4-additive-scenario-3-p1-r17: not satisfied — 11 recorded calls, and none named any of the 1 document; no call to a tool of an MCP server session L4-additive-scenario-3-p1-r18: not satisfied — 11 recorded calls, and none named any of the 1 document; no call to a tool of an MCP server session L4-additive-scenario-3-p1-r2: not satisfied — 15 recorded calls, and none named any of the 1 document; no call to a tool of an MCP server session L4-additive-scenario-3-p1-r3: not satisfied — 18 recorded calls, and none named any of the 1 document; no call to a tool of an MCP server session L4-additive-scenario-3-p1-r4: satisfied — event 13, call 8:Bash {"command":"grep -n \"summary\" taxonomy-source/headwater-standard/*.yml 2>/dev/null | head -30; echo ---; find . -iname \"taxonomy.yml\" -not -path \"*/fixtures/*\" | head; echo ---; grep -rn \"facet.value.blank\\|facet_blank\" docs/spec/12-check-layer.md"}; no call to a tool of an MCP server session L4-additive-scenario-3-p1-r5: not satisfied — 12 recorded calls, and none named any of the 1 document; no call to a tool of an MCP server session L4-additive-scenario-3-p1-r6: not satisfied — 20 recorded calls, and none named any of the 1 document; no call to a tool of an MCP server session L4-additive-scenario-3-p1-r7: not satisfied — 34 recorded calls, and none named any of the 1 document; no call to a tool of an MCP server session L4-additive-scenario-3-p1-r8: not satisfied — 15 recorded calls, and none named any of the 1 document; no call to a tool of an MCP server session L4-additive-scenario-3-p1-r9: not satisfied — 44 recorded calls, and none named any of the 1 document; no call to a tool of an MCP server - HW-PROBE-a-session-follows-two-citations-from-a-record-the-task-does-not-name (sufficiency, expects answered)
session L4-additive-scenario-3-p3-r1: satisfied — event 37: answered
HW-DR-0097; no call to a tool of an MCP server session L4-additive-scenario-3-p3-r10: satisfied — event 38: answeredHW-DR-0097; no call to a tool of an MCP server session L4-additive-scenario-3-p3-r11: satisfied — event 39: answeredHW-DR-0097; no call to a tool of an MCP server session L4-additive-scenario-3-p3-r12: satisfied — event 40: answeredHW-DR-0097; no call to a tool of an MCP server session L4-additive-scenario-3-p3-r13: satisfied — event 41: answeredHW-DR-0097; no call to a tool of an MCP server session L4-additive-scenario-3-p3-r14: satisfied — event 42: answeredHW-DR-0097; no call to a tool of an MCP server session L4-additive-scenario-3-p3-r15: satisfied — event 43: answeredHW-DR-0097; no call to a tool of an MCP server session L4-additive-scenario-3-p3-r16: satisfied — event 44: answeredHW-DR-0097; no call to a tool of an MCP server session L4-additive-scenario-3-p3-r17: satisfied — event 45: answeredHW-DR-0097; no call to a tool of an MCP server session L4-additive-scenario-3-p3-r18: satisfied — event 46: answeredHW-DR-0097; no call to a tool of an MCP server session L4-additive-scenario-3-p3-r2: satisfied — event 47: answeredHW-DR-0097; no call to a tool of an MCP server session L4-additive-scenario-3-p3-r3: satisfied — event 48: answeredHW-DR-0097; no call to a tool of an MCP server session L4-additive-scenario-3-p3-r4: satisfied — event 49: answeredHW-DR-0097; no call to a tool of an MCP server session L4-additive-scenario-3-p3-r5: satisfied — event 50: answeredHW-DR-0097; no call to a tool of an MCP server session L4-additive-scenario-3-p3-r6: satisfied — event 51: answeredHW-DR-0097; no call to a tool of an MCP server session L4-additive-scenario-3-p3-r7: satisfied — event 52: answeredHW-DR-0097; no call to a tool of an MCP server session L4-additive-scenario-3-p3-r8: satisfied — event 53: answeredHW-DR-0097; no call to a tool of an MCP server session L4-additive-scenario-3-p3-r9: satisfied — event 54: answeredHW-DR-0097; no call to a tool of an MCP server - HW-PROBE-a-session-records-which-obligation-an-evaluation-discharged (sufficiency, expects patched)
session L4-additive-scenario-3-p4-r1: not satisfied —
relation.reciprocity.missingreported over each of the 1 produced artifact; no call to a tool of an MCP server session L4-additive-scenario-3-p4-r10: satisfied — event 56:relation.reciprocity.missingreported nothing overdocs/evaluations/governs-edges-what-an-anchor-reaches-what-covers-it-when-it-ages-and-where-a-session-meets-it.md; no call to a tool of an MCP server session L4-additive-scenario-3-p4-r11: not satisfied —relation.reciprocity.missingreported over each of the 1 produced artifact; no call to a tool of an MCP server session L4-additive-scenario-3-p4-r12: satisfied — event 58:relation.reciprocity.missingreported nothing overdocs/evaluations/governs-edges-what-an-anchor-reaches-what-covers-it-when-it-ages-and-where-a-session-meets-it.md; no call to a tool of an MCP server session L4-additive-scenario-3-p4-r13: not satisfied —relation.reciprocity.missingreported over each of the 1 produced artifact; no call to a tool of an MCP server session L4-additive-scenario-3-p4-r14: satisfied — event 60:relation.reciprocity.missingreported nothing overdocs/evaluations/governs-edges-what-an-anchor-reaches-what-covers-it-when-it-ages-and-where-a-session-meets-it.md; no call to a tool of an MCP server session L4-additive-scenario-3-p4-r15: satisfied — event 61:relation.reciprocity.missingreported nothing overdocs/evaluations/governs-edges-what-an-anchor-reaches-what-covers-it-when-it-ages-and-where-a-session-meets-it.md; no call to a tool of an MCP server session L4-additive-scenario-3-p4-r16: not satisfied —relation.reciprocity.missingreported over each of the 1 produced artifact; no call to a tool of an MCP server session L4-additive-scenario-3-p4-r17: not satisfied —relation.reciprocity.missingreported over each of the 1 produced artifact; no call to a tool of an MCP server session L4-additive-scenario-3-p4-r18: satisfied — event 64:relation.reciprocity.missingreported nothing overdocs/evaluations/governs-edges-what-an-anchor-reaches-what-covers-it-when-it-ages-and-where-a-session-meets-it.md; no call to a tool of an MCP server session L4-additive-scenario-3-p4-r2: not satisfied —relation.reciprocity.missingreported over each of the 1 produced artifact; no call to a tool of an MCP server session L4-additive-scenario-3-p4-r3: not satisfied —relation.reciprocity.missingreported over each of the 1 produced artifact; no call to a tool of an MCP server session L4-additive-scenario-3-p4-r4: not satisfied —relation.reciprocity.missingreported over each of the 1 produced artifact; no call to a tool of an MCP server session L4-additive-scenario-3-p4-r5: not satisfied —relation.reciprocity.missingreported over each of the 1 produced artifact; no call to a tool of an MCP server session L4-additive-scenario-3-p4-r6: not satisfied —relation.reciprocity.missingreported over each of the 1 produced artifact; no call to a tool of an MCP server session L4-additive-scenario-3-p4-r7: satisfied — event 70:relation.reciprocity.missingreported nothing overdocs/evaluations/governs-edges-what-an-anchor-reaches-what-covers-it-when-it-ages-and-where-a-session-meets-it.md; no call to a tool of an MCP server session L4-additive-scenario-3-p4-r8: not satisfied —relation.reciprocity.missingreported over each of the 1 produced artifact; no call to a tool of an MCP server session L4-additive-scenario-3-p4-r9: satisfied — event 72:relation.reciprocity.missingreported nothing overdocs/evaluations/governs-edges-what-an-anchor-reaches-what-covers-it-when-it-ages-and-where-a-session-meets-it.md; no call to a tool of an MCP server - HW-PROBE-an-agent-reaches-the-adjudication-when-the-decision-a-tombstone-links-is-superseded (navigability, expects opened)
session L4-additive-scenario-3-p2-r1: not satisfied — 4 recorded calls, and none named any of the 1 document; no call to a tool of an MCP server
session L4-additive-scenario-3-p2-r10: not satisfied — 5 recorded calls, and none named any of the 1 document; no call to a tool of an MCP server
session L4-additive-scenario-3-p2-r11: satisfied — event 21, call 6:
Read docs/decisions/0087-no-paragraph-limit-joins-the-language-rule-because-most-paragraphs-past-six-sentences-hold-one-topic.md; no call to a tool of an MCP server session L4-additive-scenario-3-p2-r12: satisfied — event 22, call 4:Read /mnt/build/hw-1775/screening/a/ws/L4-additive-scenario-3-p2-r12/docs/decisions/0087-no-paragraph-limit-joins-the-language-rule-because-most-paragraphs-past-six-sentences-hold-one-topic.md; no call to a tool of an MCP server session L4-additive-scenario-3-p2-r13: satisfied — event 23, call 6:Read /mnt/build/hw-1775/screening/a/ws/L4-additive-scenario-3-p2-r13/docs/decisions/0087-no-paragraph-limit-joins-the-language-rule-because-most-paragraphs-past-six-sentences-hold-one-topic.md; no call to a tool of an MCP server session L4-additive-scenario-3-p2-r14: not satisfied — 6 recorded calls, and none named any of the 1 document; no call to a tool of an MCP server session L4-additive-scenario-3-p2-r15: not satisfied — 2 recorded calls, and none named any of the 1 document; no call to a tool of an MCP server session L4-additive-scenario-3-p2-r16: not satisfied — 5 recorded calls, and none named any of the 1 document; no call to a tool of an MCP server session L4-additive-scenario-3-p2-r17: satisfied — event 27, call 11:Read docs/decisions/0087-no-paragraph-limit-joins-the-language-rule-because-most-paragraphs-past-six-sentences-hold-one-topic.md; no call to a tool of an MCP server session L4-additive-scenario-3-p2-r18: satisfied — event 28, call 5:Bash {"command":"cat \"/mnt/build/hw-1775/screening/a/ws/L4-additive-scenario-3-p2-r18/docs/decisions/0087-no-paragraph-limit-joins-the-language-rule-because-most-paragraphs-past-six-sentences-hold-one-topic.md\"","description":"Read decision record 0087"}; no call to a tool of an MCP server session L4-additive-scenario-3-p2-r2: satisfied — event 29, call 4:Read /mnt/build/hw-1775/screening/a/ws/L4-additive-scenario-3-p2-r2/docs/decisions/0087-no-paragraph-limit-joins-the-language-rule-because-most-paragraphs-past-six-sentences-hold-one-topic.md; no call to a tool of an MCP server session L4-additive-scenario-3-p2-r3: satisfied — event 30, call 4:Read /mnt/build/hw-1775/screening/a/ws/L4-additive-scenario-3-p2-r3/docs/decisions/0087-no-paragraph-limit-joins-the-language-rule-because-most-paragraphs-past-six-sentences-hold-one-topic.md; no call to a tool of an MCP server session L4-additive-scenario-3-p2-r4: satisfied — event 31, call 4:Read /mnt/build/hw-1775/screening/a/ws/L4-additive-scenario-3-p2-r4/docs/decisions/0087-no-paragraph-limit-joins-the-language-rule-because-most-paragraphs-past-six-sentences-hold-one-topic.md; no call to a tool of an MCP server session L4-additive-scenario-3-p2-r5: satisfied — event 32, call 6:Read /mnt/build/hw-1775/screening/a/ws/L4-additive-scenario-3-p2-r5/docs/decisions/0087-no-paragraph-limit-joins-the-language-rule-because-most-paragraphs-past-six-sentences-hold-one-topic.md; no call to a tool of an MCP server session L4-additive-scenario-3-p2-r6: satisfied — event 33, call 4:Bash {"command":"cat \"/mnt/build/hw-1775/screening/a/ws/L4-additive-scenario-3-p2-r6/docs/decisions/0087-no-paragraph-limit-joins-the-language-rule-because-most-paragraphs-past-six-sentences-hold-one-topic.md\"","description":"Show decision record 0087"}; no call to a tool of an MCP server session L4-additive-scenario-3-p2-r7: satisfied — event 34, call 7:Read /mnt/build/hw-1775/screening/a/ws/L4-additive-scenario-3-p2-r7/docs/decisions/0087-no-paragraph-limit-joins-the-language-rule-because-most-paragraphs-past-six-sentences-hold-one-topic.md; no call to a tool of an MCP server session L4-additive-scenario-3-p2-r8: not satisfied — 1 recorded call, and none named any of the 1 document; no call to a tool of an MCP server session L4-additive-scenario-3-p2-r9: not satisfied — 1 recorded call, and none named any of the 1 document; no call to a tool of an MCP server
72 of 72 sessions that recorded their calls made no call to a tool of an MCP server. 0 sessions recorded no calls.
The rate, and the denominator it is over
40 of 72 graded sessions satisfied their expectation: 55.6%, in a 95% interval of 44.1% to 66.5%. The denominator is the graded sessions and never the selected probes. 0 sessions reached no verdict, and a session with no verdict is outside both halves of that fraction.
An interval that overlaps the previous run's is variance and one that does not is drift. This is one arm, so it estimates no effect: an efficacy claim is a comparison of two results, and the arm each one recorded is on it.