Add structured abstract consistency assistant

[KoiosSG]

Portfolio Comparison Refresh (2026-06-27)

• Current live state: open, non-draft, clean mergeability on head 1884f22; no GitHub check runs or status contexts are attached, and Algora remains Pending with Total paid $0.
• Same-issue distinction: newer PR Add data fabrication anomaly assistant #597 adds a data-fabrication anomaly assistant, while this PR remains the structured-abstract consistency assistant for source methods/results reconciliation, required section coverage, endpoint naming, negated wording, sample-size wording, certainty overclaim, limitation, and adverse-outcome consistency.
• Reviewer evidence advantage: this body and committed artifacts enumerate the post-hardening malformed/missing evidence gates, source endpoint disagreement holds, timestamp validation, and deterministic audit evidence already verified for the structured-abstract slice.

/claim #16 ## Summary Adds a distinct structured-abstract-consistency-assistant/ slice for Scientific Bounty System issue #16. The assistant evaluates structured manuscript abstracts before AI peer-review packets or editor summaries are shown. It checks required abstract sections, affirmed methods design, negated design wording, source methods/results endpoint availability and reconciliation, negated primary endpoint wording, target-specific sample-size counts, primary endpoint naming, bidirectional result direction, certainty overclaims under exploratory or null-crossing evidence, limitation language, safety/adverse-outcome wording, and deterministic audit evidence. ## Hardening Updates - Holds abstracts when source methods/results evidence packets are missing, so polished abstract text cannot release without authoritative comparison data. - Holds otherwise complete abstracts with invalid or missing assessedAt timestamps, so AI peer-review packets cannot release with unauditable structured-abstract timing evidence. - Holds ISO-looking but calendar-impossible assessedAt timestamps such as 2026-02-30T10:00:00Z, so JavaScript date normalization cannot release AI peer-review packets with shifted timing evidence. - Holds present source methods evidence packets that omit primaryEndpoint, so abstract/results alignment cannot release without a named methods endpoint anchor. - Holds present source results evidence packets that omit primaryEndpoint, so generic primary-endpoint claims cannot release without a named results evidence anchor. - Holds methods/results source endpoint disagreements, so an abstract cannot release by following only one source evidence packet. - Holds abstracts that mention the expected methods design only to deny it, such as not a retrospective cohort, so substring matches cannot release contradictory method summaries. - Holds ordinal measurement wording such as 96th percentile when it appears where manuscript/participant counts are required. - Holds hyphenated measurement wording such as 96-hour or 96-point when it appears where sample-size counts are required. - Holds duration/effect measurements such as 96 hours or 96 minutes when they masquerade as count evidence. - Holds abbreviated scientific/time units such as 96 h, 96 mg, and 96 mmHg when they masquerade as count evidence. - Holds decimal values such as 0.96 and percentage wording such as 96% unless the actual count is also stated. - Accepts normal comma-formatted counts such as 1,200 as valid count evidence. - Blocks generic primary-endpoint wording unless the named endpoint from the result packet is present. - Holds abstracts that mention the expected primary endpoint only to deny it, such as not comment triage time, so substring matches cannot release contradictory result summaries. - Blocks result and conclusion direction drift in both directions, including improvement claims over no-effect/worse evidence and worse/no-effect wording over improved evidence. - Blocks no-difference and equivalence outcome wording such as no meaningful difference, no superiority, or outcomes were comparable when the result packet records improvement. - Blocks result-section or conclusion-section certainty overclaims such as statistically significant, clinically meaningful, robust, proven, or definitive when evidence is exploratory or confidence intervals cross null. - Treats mixed phrasing such as not statistically significant but clinically meaningful as an overclaim instead of clearing it through the negated significant phrase. - Requires specific limitation language beyond weak hedges such as may when evidence is exploratory or null-crossing. - Allows accurate adverse-outcome wording when the result packet also records a worse direction, while blocking safety-benefit or negated-safety-concern conclusions over worsened adverse outcomes. ## Non-overlap This is scoped to structured abstract consistency before AI review release. It does not duplicate broad assistant suites, evidence/protocol trace modules, statistics review, research-gap planning, rebuttal packs, ethics/data review, citation context, reporting guidelines, benchmark leakage, figure/table consistency, analysis-variable provenance, domain templates, grant fit, limitations disclosure, uncertainty calibration, supplement readiness, prompt safety, study power, COI/funding, retraction, preregistration, external validity, image integrity, assay-control/calibration, literature freshness, randomization/blinding, Bayesian prior sensitivity, systematic screening drift, sample chain-of-custody, or model-assumption diagnostics slices. ## Validation - Latest impossible-calendar timestamp regression failed before implementation with release_peer_review_packet instead of hold_peer_review_packet when assessedAt was 2026-02-30T10:00:00Z. - Latest invalid-assessment-timestamp regression failed before implementation with release_peer_review_packet instead of hold_peer_review_packet when an otherwise complete abstract used assessedAt: not-a-date. - Latest missing-methods-endpoint regression failed before implementation with release_peer_review_packet instead of hold_peer_review_packet when source methods evidence omitted primaryEndpoint but the abstract/results evidence named comment triage time. - Latest source-endpoint regression failed before implementation with release_peer_review_packet instead of hold_peer_review_packet when methods evidence named reviewer workload score and results evidence named comment triage time. - Latest missing-results-endpoint regression failed before implementation with release_peer_review_packet instead of hold_peer_review_packet when source results evidence omitted primaryEndpoint but the abstract claimed the primary endpoint improved. - Latest no-difference/equivalence outcome regression failed before implementation with release_peer_review_packet instead of hold_peer_review_packet when improved result evidence was summarized as no meaningful difference / comparable outcomes. - Prior abbreviated-unit regression failed before implementation with release_peer_review_packet instead of hold_peer_review_packet when abstract methods/results used 96 h duration wording where count evidence was required. - Prior missing-source-evidence regression failed before implementation with release_peer_review_packet instead of hold_peer_review_packet when a complete abstract lacked source methods/results evidence packets. - Prior negated-endpoint regression failed before implementation with release_peer_review_packet instead of hold_peer_review_packet when an abstract stated not comment triage time while the result packet required comment triage time. - Prior negated-design regression failed before implementation with release_peer_review_packet instead of hold_peer_review_packet when an abstract stated not a retrospective cohort while the method packet required retrospective cohort. - npm test from structured-abstract-consistency-assistant -> structured-abstract-consistency-assistant tests passed (34). - npm run check -> syntax checks passed for index, sample data, test, and demo files. - npm run demo -> regenerated 23 JSON packets plus Markdown/SVG evidence, including invalid-assessed-at-packet.json and malformed-manuscript-packet.json. - npm run video -> regenerated reports/demo.mp4. - All 23 generated JSON packets parsed successfully. - ffprobe verified reports/demo.mp4 as H.264, 1280x720, 24 fps, 7.5s, 115,991 bytes. - git diff --check and git diff --cached --check passed; only Windows line-ending normalization warnings appeared before staging. - Focused restricted-string scan returned no credential, payout, or token strings. - GitHub PR merge state after push: CLEAN; no checks are reported for this branch. ## Demo Artifacts - structured-abstract-consistency-assistant/reports/blocked-packet.json - structured-abstract-consistency-assistant/reports/revision-packet.json - structured-abstract-consistency-assistant/reports/negated-design-packet.json - structured-abstract-consistency-assistant/reports/negated-primary-endpoint-packet.json - structured-abstract-consistency-assistant/reports/missing-source-evidence-packet.json - structured-abstract-consistency-assistant/reports/result-certainty-packet.json - structured-abstract-consistency-assistant/reports/mixed-certainty-packet.json - structured-abstract-consistency-assistant/reports/conclusion-certainty-packet.json - structured-abstract-consistency-assistant/reports/weak-limitation-packet.json - structured-abstract-consistency-assistant/reports/percentage-sample-size-packet.json - structured-abstract-consistency-assistant/reports/decimal-sample-size-packet.json - structured-abstract-consistency-assistant/reports/duration-sample-size-packet.json - structured-abstract-consistency-assistant/reports/abbreviated-unit-sample-size-packet.json - structured-abstract-consistency-assistant/reports/hyphenated-measurement-sample-size-packet.json - structured-abstract-consistency-assistant/reports/ordinal-sample-size-packet.json - structured-abstract-consistency-assistant/reports/no-difference-outcome-packet.json - structured-abstract-consistency-assistant/reports/missing-results-endpoint-packet.json - structured-abstract-consistency-assistant/reports/missing-methods-endpoint-packet.json - structured-abstract-consistency-assistant/reports/source-endpoint-mismatch-packet.json - structured-abstract-consistency-assistant/reports/invalid-assessed-at-packet.json - structured-abstract-consistency-assistant/reports/impossible-assessed-at-packet.json - structured-abstract-consistency-assistant/reports/clean-packet.json - structured-abstract-consistency-assistant/reports/abstract-consistency-report.md - structured-abstract-consistency-assistant/reports/summary.svg - structured-abstract-consistency-assistant/reports/demo.mp4 Synthetic data only. No external services, credentials, live databases, private manuscripts, or payment data are used. AI-assisted with OpenAI Codex; I reviewed and locally verified the diff before submitting.

[KoiosSG]

Hardening update pushed in 4ac8b8c: preserved same-code findings for different evidence targets, so methods and results sample-size mismatches are both reported instead of one being collapsed. Validation refreshed locally: npm run check, npm test (4 tests), npm run demo, npm run video, ffprobe on reports/demo.mp4, git diff --check, and a sensitive-term scan with no matches.

[KoiosSG]

Hardening update pushed in 5e04a92: structured abstract results now must name the actual primary endpoint instead of passing on generic primary endpoint wording. This closes a reviewer-readiness gap where an abstract could claim a primary endpoint improved without identifying the manuscript endpoint being summarized.

I added a regression that failed before the fix with release_peer_review_packet == hold_peer_review_packet for generic endpoint language and now passes. The existing blocked demo packet now also records ENDPOINT_MISMATCH, and the generated Markdown/JSON/SVG artifacts were refreshed.

Validation refreshed locally:

• npm test -> structured-abstract-consistency-assistant tests passed (5)
• npm run check -> syntax checks passed for index, sample-data, test, and demo
• npm run demo -> regenerated blocked/revision/clean packets and report artifacts
• npm run video -> regenerated demo.mp4
• ffprobe on reports/demo.mp4 -> H.264, 1280x720, 7.5s, 24fps, 109,710 bytes
• git diff --check and git diff --cached --check
• sensitive-term scan with rg -n "(password|secret|wallet|paypal|bank|passport|private key|api key)" structured-abstract-consistency-assistant returned no matches

[KoiosSG]

Follow-up competitive hardening pass for the structured abstract consistency assistant.

What changed:

• Added a regression for exploratory/uncertain evidence where limitation text exists elsewhere in the manuscript but the structured abstract conclusion itself lacks limitation language.
• Tightened the release gate so reviewer-facing structured abstracts must carry limitation language in the conclusion before AI peer-review/editor packets are released.
• This makes the slice more defensible in a crowded issue AI-Powered Research Assistant Suite #16 field because it protects the exact user-facing abstract output, not just the presence of limitation metadata somewhere else in the packet.

Validation:

• Confirmed the new regression failed before the implementation: the packet was incorrectly release_peer_review_packet instead of hold_peer_review_packet.
• npm test -> 6 structured abstract consistency assistant tests passed.
• npm run check -> JS syntax checks passed.
• npm run demo -> generated blocked/revision/clean packets with expected statuses.
• npm run video -> demo video generation passed.
• ffprobe verified reports/demo.mp4 as H.264, 1280x720, 24 fps, 7.5s, 109,710 bytes.
• git diff --check and git diff --cached --check passed; the only messages were Git line-ending normalization warnings on Windows.
• Sensitive-term scan of the code/test patch found no payout or credential strings.

[KoiosSG]

Follow-up competitive hardening pass for the structured abstract consistency assistant.

What changed in 8608676:

• Added a regression for a reviewer-facing risk where a structured abstract described the primary endpoint as improved even though the result packet recorded a worse/harmful direction.
• Tightened result-direction gating so abstracts with improvement language are held when the underlying result direction is no_clear_effect, worse, harmful, inferior, declined, or negative.
• Kept the reviewer-facing message polished by rendering internal direction tokens as readable text in generated findings.

Validation refreshed locally:

• Confirmed the new regression failed before implementation: the packet was incorrectly release_peer_review_packet instead of hold_peer_review_packet.
• npm test -> 7 structured abstract consistency assistant tests passed.
• npm run check -> JS syntax checks passed.
• npm run demo -> blocked/revision/clean packets regenerated with expected statuses.
• npm run video -> demo video generation passed.
• ffprobe verified reports/demo.mp4 as H.264, 1280x720, 24 fps, 7.5s, 109,710 bytes.
• git diff --check and git diff --cached --check passed; only Git line-ending normalization warnings appeared on Windows.
• Sensitive-term scan of the assistant returned no payout or credential strings.

[KoiosSG]

Follow-up competitive hardening pass for the structured abstract consistency assistant.

What changed in 8388acc:

• Added a regression for a reviewer-facing risk where the structured abstract results accurately report a worse direction, but the conclusion still claims the assistant is effective or beneficial.
• Tightened conclusion-direction gating so AI peer-review/editor packets are held when the conclusion implies benefit while the result packet records no_clear_effect, worse, harmful, inferior, declined, or negative outcomes.
• Refreshed the generated Markdown/JSON/SVG evidence so the blocked packet now includes CONCLUSION_RESULT_DIRECTION_MISMATCH.

Validation refreshed locally:

• Confirmed the new regression failed before implementation: the packet was incorrectly release_peer_review_packet instead of hold_peer_review_packet.
• npm test -> 8 structured abstract consistency assistant tests passed.
• npm run demo -> regenerated blocked/revision/clean packets with expected statuses.
• npm run video -> regenerated demo.mp4.
• npm run check -> JS syntax checks passed for index, sample-data, test, and demo.
• ffprobe verified reports/demo.mp4 as H.264, 1280x720, 24 fps, 7.5s, 109,710 bytes.
• git diff --check and git diff --cached --check passed; only Git line-ending normalization warnings appeared on Windows.
• Sensitive-term scan of the assistant returned no payout or credential strings.

[KoiosSG]

Follow-up competitive hardening pass for the structured abstract consistency assistant.

What changed in a80df5d:

• Added a regression for reviewer-ready abstracts that express sample sizes with normal comma formatting, e.g. 1,200, while manuscript evidence stores the numeric count as 1200.
• Sample-size evidence matching now accepts comma-formatted integer counts in methods/results abstract text without weakening mismatch detection.
• README, requirements map, and acceptance notes now explicitly cover formatted manuscript counts.

Validation refreshed locally:

• Confirmed the new regression failed before implementation with hold_peer_review_packet instead of release_peer_review_packet.
• npm test -> 9 structured abstract consistency assistant tests passed.
• npm run demo -> regenerated blocked/revision/clean packets with expected statuses.
• npm run video -> regenerated demo.mp4.
• npm run check -> syntax checks passed for index, sample-data, test, and demo.
• ffprobe verified reports/demo.mp4 as H.264, 1280x720, 24 fps, 7.5s, 109,710 bytes.
• git diff --check and git diff --cached --check passed; only Git line-ending normalization warnings appeared on Windows.
• Sensitive-term scan returned no payout or credential strings.
• GitHub PR merge state after push: CLEAN.

[KoiosSG]

Follow-up competitive hardening pass for the structured abstract consistency assistant.

What changed in fcac586:

• Added a regression for lower-outcome benefit wording where an abstract said the primary endpoint was lower even though the result packet recorded no_clear_effect.
• Expanded result/conclusion benefit-language detection beyond improved/reduced phrasing to include lower, lowered, shorter, faster, improvement, and reduction wording.
• The assistant now blocks both result-section and conclusion-section lower-outcome claims when result direction is no clear effect, worse, harmful, inferior, declined, or negative.
• README, requirements map, and acceptance notes now explicitly cover lower/shorter/faster benefit wording.

Validation refreshed locally:

• Confirmed the new regression failed before implementation with release_peer_review_packet instead of hold_peer_review_packet.
• npm test -> structured-abstract-consistency-assistant tests passed (10).
• npm run demo -> regenerated blocked/revision/clean packets with expected statuses.
• npm run video -> regenerated reports/demo.mp4.
• npm run check -> syntax checks passed for index, sample-data, test, and demo.
• ffprobe verified reports/demo.mp4 as H.264, 1280x720, 24 fps, 7.5s, 109,710 bytes.
• git diff --check and git diff --cached --check passed; only Git line-ending normalization warnings appeared on Windows before staging.
• Sensitive-term scan returned no payout or credential strings.
• GitHub PR merge state after push: CLEAN.

[KoiosSG]

Follow-up competitive hardening pass for the structured abstract consistency assistant.

What changed in 06d0c03:

• Added a regression for a reviewer-facing false blocker where an abstract accurately says an adverse outcome increased and the result packet also records a worse direction.
• Benefit-language detection no longer treats bare increase / increased as an improvement claim when the text is clearly about adverse outcomes such as adverse events, mortality, harms, complications, toxicity, failures, errors, or infections.
• Positive increased claims still remain guarded when they are not adverse-outcome wording, so the change reduces false blockers without weakening no-effect/worse-result drift checks.
• README, requirements map, and acceptance notes now explicitly cover accurate adverse-outcome wording.

Validation refreshed locally:

• Confirmed the new regression failed before implementation with hold_peer_review_packet instead of release_peer_review_packet.
• npm test -> structured-abstract-consistency-assistant tests passed (11).
• npm run demo -> regenerated blocked/revision/clean packets with expected statuses.
• npm run video -> regenerated reports/demo.mp4.
• npm run check -> syntax checks passed for index, sample-data, test, and demo.
• ffprobe verified reports/demo.mp4 as H.264, 1280x720, 24 fps, 7.5s, 109,710 bytes.
• git diff --check and git diff --cached --check passed; only Git line-ending normalization warnings appeared on Windows.
• Sensitive-term scan returned no payout or credential strings.
• GitHub PR merge state after push: CLEAN.

[KoiosSG]

Follow-up competitive hardening pass for the structured abstract consistency assistant.

What changed in f9c6ffd:

• Added a regression for the opposite result-direction drift case: the structured abstract says the primary endpoint worsened while the result packet records improvement.
• Result-direction checks are now bidirectional. The assistant blocks over-positive wording for no-effect/worse evidence and also blocks worsened, harmful, or no-effect wording when the packet records improvement.
• The mismatch messages and reviewer artifacts were refreshed so the contract is stated as directional evidence alignment, not only benefit-claim policing.
• README, requirements map, acceptance notes, blocked packet, and summary visual now explicitly cover bidirectional result-direction drift.

Validation refreshed locally:

• Confirmed the new regression failed before implementation with release_peer_review_packet instead of hold_peer_review_packet.
• npm test -> structured-abstract-consistency-assistant tests passed (12).
• npm run demo -> regenerated blocked/revision/clean packets with expected statuses.
• npm run video -> regenerated reports/demo.mp4.
• npm run check -> syntax checks passed for index, sample-data, test, and demo.
• node --check passed for index, sample-data, test, and demo.
• ffprobe verified reports/demo.mp4 as H.264, 1280x720, 24 fps, 7.5s, 109,710 bytes.
• git diff --check and git diff --cached --check passed; only Git line-ending normalization warnings appeared on Windows.
• Sensitive-term scan returned no payout or credential strings.
• GitHub PR merge state after push: CLEAN.

[KoiosSG]

Follow-up competitive hardening pass for the structured abstract consistency assistant.

What changed in e0583ce:

• Added a regression for a reviewer-facing safety overclaim: a structured abstract conclusion says an assistant was safe and well tolerated while the result packet records worsened adverse-outcome evidence.
• Safety-benefit wording is now treated as positive conclusion language unless it is explicitly negated, so adverse-outcome evidence can block misleading safety conclusions before AI peer-review/editor packets are released.
• Negative safety wording such as not safe, unsafe, poorly tolerated, or safety concerns is treated as adverse/worse-outcome language rather than as a false positive safety benefit.
• README, requirements map, and acceptance notes now explicitly cover safety-benefit conclusion drift over worse adverse-outcome evidence.

Validation refreshed locally:

• Confirmed the new regression failed before implementation with release_peer_review_packet instead of hold_peer_review_packet.
• npm test -> structured-abstract-consistency-assistant tests passed (13).
• npm run demo -> regenerated blocked/revision/clean packets with expected statuses.
• npm run video -> regenerated reports/demo.mp4.
• npm run check -> syntax checks passed for index, sample-data, test, and demo.
• ffprobe verified reports/demo.mp4 as H.264, 1280x720, 24 fps, 7.5s, 109,710 bytes.
• git diff --check and git diff --cached --check passed; only Git line-ending normalization warnings appeared on Windows.
• Sensitive-term scan returned no payout or credential strings.
• GitHub PR merge state after push: CLEAN.

[KoiosSG]

Follow-up competitive hardening pass for the structured abstract consistency assistant.

What changed in 18db72d:

• Added a regression for negated benefit wording where a structured abstract says the primary endpoint did not improve while the result packet records improvement.
• Negated benefit phrases are now treated as no-effect/worse wording instead of accidentally matching the positive improve token.
• Both result-section and conclusion-section negated benefit drift now hold the AI peer-review/editor packet before release.
• README, requirements map, and acceptance notes now make negated benefit wording part of the reviewer-visible evidence-alignment contract.

Validation refreshed locally:

• Confirmed the new regression failed before implementation with release_peer_review_packet instead of hold_peer_review_packet.
• npm test -> structured-abstract-consistency-assistant tests passed (14).
• npm run demo -> regenerated blocked/revision/clean packets with expected statuses.
• npm run video -> regenerated reports/demo.mp4.
• npm run check -> syntax checks passed for index, sample-data, test, and demo.
• node --check passed for index, sample-data, test, and demo.
• ffprobe verified reports/demo.mp4 as H.264, 1280x720, 24 fps, 7.5s, 109,710 bytes.
• git diff --check and git diff --cached --check passed.
• Sensitive-term scan returned no payout or credential strings.

[KoiosSG]

Follow-up competitive hardening pass for the structured abstract consistency assistant.

What changed in 320404d:

• Added a regression for negated safety-concern conclusions where a structured abstract says no safety concerns while the result packet records worsened adverse-outcome evidence.
• The assistant now treats phrases such as no safety concerns / no adverse events as safety-benefit claims, not as adverse/worse-outcome language.
• Such conclusions now hold the AI peer-review/editor packet with CONCLUSION_RESULT_DIRECTION_MISMATCH until the abstract matches the adverse-event evidence.
• README, requirements map, and acceptance notes now include this negated safety-concern contract.

Why this matters:

• Safety language is a high-review-risk part of issue AI-Powered Research Assistant Suite #16. A conclusion that says no safety concern while evidence worsens is materially misleading even though it contains the word concerns.
• This keeps PR Add structured abstract consistency assistant #419 ahead on reviewer-facing abstract/evidence alignment without broadening into adjacent issue AI-Powered Research Assistant Suite #16 submissions.

Validation refreshed locally:

• Confirmed the new regression failed before implementation with release_peer_review_packet instead of hold_peer_review_packet.
• npm test -> structured-abstract-consistency-assistant tests passed (15).
• npm run check -> syntax checks passed for index, sample-data, test, and demo.
• npm run demo -> regenerated blocked/revision/clean packets with expected statuses.
• npm run video -> regenerated reports/demo.mp4.
• ffprobe verified reports/demo.mp4 as H.264, 1280x720, 24 fps, 7.5s, 109,710 bytes.
• git diff --check and git diff --cached --check passed; only Git line-ending normalization warnings appeared on Windows.
• Sensitive-term scan returned no payout or credential strings.
• GitHub PR merge state after push: CLEAN; no checks are reported for this branch.

[KoiosSG]

Follow-up competitive hardening pass for the structured abstract consistency assistant.

What changed in bc627d5:

• Added a regression for structured abstract Results text that claims a statistically significant and clinically meaningful improvement while the evidence crosses null.
• The assistant now emits RESULT_OVERSTATES_EVIDENCE and holds AI peer-review/editor packets when Results-section certainty language outruns exploratory or null-crossing evidence.
• Added reports/result-certainty-packet.json and refreshed the Markdown/SVG/MP4 reviewer evidence so the new gate is visible in the PR artifacts.
• README, requirements map, and acceptance notes now explicitly cover result-section certainty overclaims.

Why this matters:

• The previous gates handled overconfident conclusions, but an abstract can mislead reviewers just as directly by putting unsupported certainty in Results.
• This keeps PR Add structured abstract consistency assistant #419 tightly focused on structured-abstract/evidence alignment while covering a high-review-risk scientific reporting failure mode.

Validation refreshed locally:

• Confirmed the new regression failed before implementation with release_peer_review_packet instead of hold_peer_review_packet.
• npm test -> structured-abstract-consistency-assistant tests passed (16).
• npm run check -> syntax checks passed for index, sample-data, test, and demo.
• npm run demo -> regenerated blocked/revision/result-certainty/clean packets with expected statuses.
• npm run video -> regenerated reports/demo.mp4.
• ffprobe verified reports/demo.mp4 as H.264, 1280x720, 24 fps, 7.5s, 109,882 bytes.
• git diff --check and git diff --cached --check passed; only Git line-ending normalization warnings appeared on Windows before staging.
• Sensitive-term scan returned no payout or credential strings.
• GitHub PR merge state after push: CLEAN; no checks are reported for this branch.

[KoiosSG]

Follow-up competitive hardening pass for the structured abstract consistency assistant.

What changed in a38417e:

• Added a regression for abstract text that mentions the expected sample-size number only inside hyphenated measurement wording, e.g. 96-hour in Methods and 96-point in Results.
• The sample-size matcher now rejects hyphenated measurement units such as hours, points, scores, ratios, odds, hazards, confidence, and CI as count evidence.
• Added reports/hyphenated-measurement-sample-size-packet.json and refreshed Markdown/SVG/MP4 reviewer evidence so this gate is visible alongside the existing percentage, decimal, and duration sample-size guards.
• README, requirements map, acceptance notes, and the PR body now explicitly cover hyphenated measurement false passes.

Why this matters:

• Scientific abstracts often hyphenate measurement modifiers. Treating 96-hour or 96-point as proof of n=96 can let an abstract look sample-size aligned while hiding that no actual manuscript or participant count was stated.
• This keeps PR Add structured abstract consistency assistant #419 focused on structured-abstract/evidence alignment while strengthening the largest active claim against a realistic reviewer-facing false pass.

Validation refreshed locally:

• Confirmed the new regression failed before implementation with release_peer_review_packet instead of hold_peer_review_packet.
• npm test -> structured-abstract-consistency-assistant tests passed (23).
• npm run demo -> regenerated reviewer packets including hyphenated-measurement-sample-size-packet.json, which holds with two SAMPLE_SIZE_MISMATCH findings.
• npm run video -> regenerated reports/demo.mp4.
• npm run check -> syntax checks passed for index, sample-data, test, and demo.
• ffprobe verified reports/demo.mp4 as H.264, 1280x720, 24 fps, 7.5s, 115,708 bytes.
• git diff --check and git diff --cached --check passed; only Windows line-ending normalization warnings appeared.
• Focused sensitive scan returned no payout, credential, or token strings.
• GitHub PR merge state after push: CLEAN; no checks are reported for this branch.

[KoiosSG]

Follow-up competitive hardening pass for the structured abstract consistency assistant.

What changed in 721c0a0:

• Added a regression for abstract text that mentions the expected sample-size number only as an ordinal measurement, e.g. 96th percentile in Methods and Results.
• The sample-size matcher now rejects ordinal suffixes (st, nd, rd, th) as count evidence, so abstracts must state the actual manuscript/participant count before AI peer-review/editor packets are released.
• Added reports/ordinal-sample-size-packet.json and refreshed the Markdown/SVG reviewer evidence so the new gate is visible alongside the percentage, decimal, duration, and hyphenated measurement guards.
• README, requirements map, acceptance notes, and the PR body now explicitly cover ordinal measurement false passes.

Why this matters:

• Percentiles and ordinal measurements are common in scientific abstracts. Treating 96th percentile as proof of n=96 can let an abstract look sample-size aligned while hiding that no actual sample size was stated.
• This keeps PR Add structured abstract consistency assistant #419 focused on structured-abstract/evidence alignment while tightening another realistic reviewer-facing false pass in the largest active claim.

Validation refreshed locally:

• Confirmed the new regression failed before implementation with release_peer_review_packet instead of hold_peer_review_packet.
• npm test -> structured-abstract-consistency-assistant tests passed (24).
• npm run demo -> regenerated reviewer packets including ordinal-sample-size-packet.json, which holds with two SAMPLE_SIZE_MISMATCH findings.
• npm run video -> regenerated reports/demo.mp4.
• npm run check -> syntax checks passed for index, sample-data, test, and demo.
• Full local sequence npm test; npm run demo; npm run video; npm run check passed.
• ffprobe verified reports/demo.mp4 as H.264, 1280x720, 24 fps, 7.5s, 115,708 bytes.
• All 12 generated JSON packets parsed successfully.
• git diff --check and git diff --cached --check passed; only Windows line-ending normalization warnings appeared before staging.
• Focused sensitive scan returned no payout, credential, or token strings.
• GitHub PR merge state after push: CLEAN; no checks are reported for this branch.

[KoiosSG]

Follow-up competitive hardening pass for the structured abstract consistency assistant.

What changed in c02b7e3:

• Added a regression for abstracts that mention the expected methods design only to deny it, e.g. not a retrospective cohort while the method packet requires retrospective cohort.
• The design-alignment check now treats immediate negation around the expected design as METHODS_DESIGN_MISMATCH instead of accepting the substring match.
• The affected abstract is held for AI peer-review release with revise_methods_summary remediation.
• Added reports/negated-design-packet.json and refreshed the Markdown/SVG/MP4 evidence so reviewers can inspect the new guard directly.
• README, requirements map, acceptance notes, sample data, and demo outputs now document the negated-design case.

Why this matters:

• Abstract consistency needs affirmative method-design evidence. Without this, an abstract could explicitly contradict the methods packet while still passing because the expected design phrase appeared inside a denial.
• This keeps PR Add structured abstract consistency assistant #419 focused on its distinct structured-abstract review-readiness slice and closes a concrete false-release path in the active issue AI-Powered Research Assistant Suite #16 competition.

Validation refreshed locally:

• Confirmed the new regression failed before implementation with release_peer_review_packet instead of hold_peer_review_packet.
• npm test -> structured-abstract-consistency-assistant tests passed (25).
• npm run demo -> regenerated 13 JSON packets plus Markdown/SVG evidence, including negated-design-packet.json.
• npm run video -> regenerated reports/demo.mp4.
• npm run check -> syntax checks passed for index, sample data, test, and demo files.
• ffprobe verified reports/demo.mp4 as H.264, 1280x720, 24 fps, 7.5s, 115,991 bytes.
• All 13 generated JSON packets parsed successfully.
• git diff --check and git diff --cached --check passed; only Windows line-ending normalization warnings appeared before staging.
• Focused sensitive scan returned no payout, credential, or token strings.
• GitHub PR merge state after push: CLEAN; no checks are reported for this branch.

[KoiosSG]

Pushed a targeted hardening update in 534329b.

What changed:

• Added a regression for abstracts that mention the expected primary endpoint only to deny it, e.g. not comment triage time.
• Reused the negation-aware phrase check for results.primaryEndpoint, so substring matches no longer release contradictory result summaries.
• Added negated-primary-endpoint-packet.json plus regenerated Markdown/SVG demo evidence.

Fresh validation:

• npm test -> structured-abstract-consistency-assistant tests passed (26)
• npm run demo -> regenerated 14 JSON packets plus Markdown/SVG evidence
• npm run video -> regenerated reports/demo.mp4
• npm run check -> syntax checks passed
• ffprobe -> H.264, 1280x720, 7.5s, 180 frames
• all generated JSON packets parsed successfully
• git diff --check and git diff --cached --check passed
• focused restricted-string scan returned no matches

[KoiosSG]

Hardening update pushed in 766c9f7 for missing source methods/results evidence.

What changed:

• Added a regression for an otherwise polished structured abstract where the source methods and results evidence packets are absent.
• The assistant now holds release with MISSING_METHODS_EVIDENCE and MISSING_RESULTS_EVIDENCE instead of treating unverified abstract text as reviewer-ready.
• Remediation now routes to attach_source_evidence, and the demo emits reports/missing-source-evidence-packet.json.

Why this matters:

• Structured abstract consistency is only meaningful when the abstract is compared against authoritative methods/results evidence. Without this gate, a polished abstract could reach AI peer-review/editor output with no source packet behind it.

Validation refreshed locally:

• Confirmed the new regression failed before implementation with release_peer_review_packet instead of hold_peer_review_packet.
• npm test -> structured-abstract-consistency-assistant tests passed (27).
• npm run demo -> regenerated 15 JSON packets plus Markdown/SVG evidence.
• npm run video -> regenerated reports/demo.mp4.
• npm run check -> syntax checks passed for index, sample-data, test, and demo.
• node --check passed for index, sample-data, test, and demo.
• ffprobe verified reports/demo.mp4 as H.264, 1280x720, 24 fps, 7.5s, 180 frames.
• All generated JSON packets parsed successfully, including missing-source-evidence-packet.json.
• git diff --check and git diff --cached --check passed; only Windows line-ending normalization warnings appeared before staging.
• Focused restricted-string scan returned no matches.

[KoiosSG]

Hardening update pushed in 7375c03 for abbreviated measurement units in structured abstract sample-size evidence.

What changed:

• Added a red regression for abstracts that use 96 h duration wording where manuscript/participant count evidence is required.
• The sample-size guard now rejects short scientific/time units such as h, hr, mg, mmHg, mL, and related units instead of accepting them as counts.
• Added reports/abbreviated-unit-sample-size-packet.json and refreshed Markdown/SVG evidence.

Fresh validation:

• npm test -> structured-abstract-consistency-assistant tests passed (28)
• npm run demo -> regenerated 16 JSON packets plus Markdown/SVG evidence
• npm run video -> regenerated reports/demo.mp4
• npm run check -> syntax checks passed
• all 16 generated JSON packets parsed successfully
• ffprobe -> H.264, 1280x720, 24 fps, 7.5s, 180 frames
• git diff --check and git diff --cached --check passed
• focused secret/payout scan returned no matches
• GitHub PR merge state after push: CLEAN; no checks are reported for this branch

[KoiosSG]

Hardening update pushed in ad86f0f: structured abstracts now hold peer-review release when improved result evidence is summarized as no-difference/equivalence wording such as no meaningful difference, no superiority, or comparable outcomes.

Validation refreshed locally:

• red regression first reproduced release_peer_review_packet instead of hold_peer_review_packet
• npm test -> structured-abstract-consistency-assistant tests passed (29)
• npm run check, npm run demo, npm run video
• parsed all 17 generated JSON packets, including no-difference-outcome-packet.json
• ffprobe verified reports/demo.mp4 as H.264 1280x720, 24fps, 7.5s, 115,991 bytes
• git diff --check, git diff --cached --check, staged allowlist check, and restricted-string scan passed

[KoiosSG]

Hardening update pushed in f42154e: source results evidence that omits primaryEndpoint now blocks structured-abstract release with MISSING_RESULTS_ENDPOINT and attach_source_evidence:*, so generic primary-endpoint wording cannot become reviewer-facing AI evidence without a named endpoint anchor.

Fresh validation from structured-abstract-consistency-assistant/:

• red regression first reproduced release_peer_review_packet instead of hold_peer_review_packet
• npm test -> 30 tests passed
• npm run check, npm run demo, and npm run video
• parsed all 18 generated JSON packets, including missing-results-endpoint-packet.json
• ffprobe verified reports/demo.mp4 as H.264 1280x720, 24fps, 7.5s, 115,991 bytes
• git diff --check, git diff --cached --check, staged allowlist check, and focused restricted-string scan passed
• GitHub PR merge state before push was CLEAN; no checks were reported for this branch

[KoiosSG]

Pushed a focused hardening commit for source endpoint reconciliation: e34dbb2.

New regression now blocks release when methods evidence and results evidence name different primary endpoints, emits SOURCE_ENDPOINT_MISMATCH, routes reconcile_source_endpoints:*, and marks both methods/results alignment signals false. Verification passed: red regression captured first, npm test (31), npm run check, npm run demo, npm run video, 19 JSON packet parses, ffprobe H.264 1280x720 24fps 7.5s, diff checks, staged allowlist check, and focused restricted-string scan. PR body has the refreshed evidence list.

[KoiosSG]

Pushed focused hardening commit ce6eb35 for missing source methods primary-endpoint evidence. New regression now holds release with MISSING_METHODS_ENDPOINT when methods evidence omits primaryEndpoint while results/abstract evidence names the endpoint, routes attach_source_evidence, and marks methodsAligned=false. Verification passed: red regression captured first, npm test (32), npm run check, npm run demo, npm run video, 20 JSON packet parses including missing-methods-endpoint-packet.json, ffprobe H.264 1280x720 24fps 7.5s, diff checks, staged allowlist check, and focused restricted-string scan. PR body has the refreshed evidence list.

[KoiosSG]

Pushed focused hardening commit 9f838f0 for a malformed top-level structured-abstract packet gap.

Before the patch, assessStructuredAbstract(null) crashed with TypeError: Cannot read properties of null (reading 'methods') before producing reviewer evidence. The package now normalizes malformed top-level input to stable unknown-manuscript evidence, emits MALFORMED_MANUSCRIPT_PACKET, keeps source-evidence and missing-section remediation, and writes reports/malformed-manuscript-packet.json.

Verified locally after the TDD regression:

• npm test -> 33 tests passed
• npm run check
• npm run demo -> 21 JSON packets regenerated and parsed
• npm run video; ffprobe confirmed H.264 1280x720, 24 fps, 7.5s, 115,991 bytes
• git diff --check, git diff --cached --check, staged allowlist check, and restricted-string scan

This is a cooldown/comment-budget exception because the gap was a newly documented high-severity crash before AI peer-review gating.

[KoiosSG]

Hardening update pushed in 4c3856d: structured abstract packets now hold AI peer-review release when assessedAt is missing or invalid, emitting INVALID_ASSESSMENT_TIMESTAMP plus repair_assessment_timestamp:* instead of releasing unauditable timing evidence.

Fresh validation from structured-abstract-consistency-assistant/:

• red regression first reproduced release_peer_review_packet instead of hold_peer_review_packet
• npm test -> 34 tests passed
• npm run check, npm run demo, npm run video
• 22 JSON packet parses, including reports/invalid-assessed-at-packet.json
• ffprobe H.264 1280x720 24fps 7.5s / 180 frames / 115,991 bytes
• git diff --check, git diff --cached --check, staged allowlist check, and focused restricted-string scan passed

[KoiosSG]

Hi SCIBASE maintainers — quick check-in on this bounty claim. This PR is still ready from my side: the latest hardening keeps the structured-abstract release gate conservative around invalid assessment timestamps, and the thread has the focused test/demo evidence.

Is there anything specific you'd like changed, simplified, or clarified to make review/selection easier? Happy to adjust quickly. Thanks!

[KoiosSG]

Hardened structured-abstract assessment timestamps in 867f844.

Fresh gap closed:

• assessedAt: "2026-02-30T10:00:00Z" previously released release_peer_review_packet because JavaScript normalized the impossible date to March 2.
• Timestamp validation now requires a strict UTC ISO timestamp whose parsed calendar components round-trip exactly, so impossible calendar dates emit INVALID_ASSESSMENT_TIMESTAMP and repair_assessment_timestamp:* before AI peer-review/editor packets can release.
• Added reports/impossible-assessed-at-packet.json plus refreshed Markdown/SVG/MP4 evidence.

Verification refreshed from structured-abstract-consistency-assistant/:

• Red regression first reproduced release_peer_review_packet instead of hold_peer_review_packet.
• npm test passed (35 tests).
• npm run check, npm run demo, and npm run video passed.
• 23 generated JSON packets parsed successfully.
• ffprobe verified reports/demo.mp4 as H.264 1280x720 24fps 7.5s / 180 frames / 115,991 bytes.
• git diff --check, git diff --cached --check, allowlist staging, and focused restricted-string scan passed.

[KoiosSG]

Quick refresh on the current hardened head for this PR: the structured abstract consistency assistant is still ready from my side.

The package gates AI peer-review/editor packet release on source evidence, endpoint and sample-size alignment, result/conclusion direction, limitation language, and strict assessment timestamp validity. Current local validation from the package directory: node test.js passes with 37 tests.

I am leaving the head stable unless maintainers would prefer a narrower shape or specific clarification for review.