Ready to supercharge your AI agents?

Install any skill in seconds. Free to start, no credit card required.

Signalling Question Answering · yogsoth-ai | DecimalAI

yogsoth-ai/signalling-question-answering

SkillScore23Maintained

Answer per-domain signalling questions (5-value scale: Yes/Probably yes/Probably no/No/No information) for RoB2, ROBINS-I, or QUADAS-2, per whichever variant study-design-tool-gate dispatched to. Use this after study-design-tool-gate has dispatched to one of these three tools; this SOP produces only the raw signalling answers, not any domain-level or overall roll-up — that happens in domain-level-judgment next.

researchby yogsoth-ai

Links this skill to your agents and keeps it current with the author’s updates.

Automatic

Clinical / Healthcare pack

1 · One-time — install the CLI (docs ↗)

pip install decimalai

2 · Get the skill — writes .claude/skills/yogsoth-ai-signalling-question-answering/SKILL.md

decimalai skills pull yogsoth-ai-signalling-question-answering --out .claude/skills/

No CLI? Get the same file — curl · raw file

curl -fsSL --create-dirs -o .claude/skills/yogsoth-ai-signalling-question-answering/SKILL.md https://app.decimal.ai/s/yogsoth-ai-signalling-question-answering/SKILL.md

Raw file: /s/yogsoth-ai-signalling-question-answering/SKILL.md ↗ · machine-readable summary at /s/yogsoth-ai-signalling-question-answering.json ↗

In productionGlobal

how this skill behaves on real traffic · last 30 days

No production data for this skill yet — usage numbers appear here once agents start running it.

Anonymized aggregate across every org using this skill — proof from real production use, not a demo.

Does this skill help? · Benchmark: with vs. withoutsign in to run this on your agents

22 test cases from the publisher’s eval.yaml · each run twice on gemini-3.6-flash — with the skill injected, then without · last run 13d ago

Yes — +5% lift+18% tokens0% turns23% of whole cases pass with skill (5/22)

Test case	Without → With	Effect	Δ tokens	Δ turns
case-06	✗→✓	▲ Improved	16%	0%
case-05	✓→✓	＝ Same ✓	40%	0%
case-19	✓→✓	＝ Same ✓	-48%	0%
case-20	✓→✓	＝ Same ✓	-23%	0%
case-21	✓→✓	＝ Same ✓	49%	0%

Showing 5 of 22 cases

SkillSafety

✓ 3 of 3 passed

Code scan15 of 15 checks clean

0 critical · 0 warning · 0 info — scanner v4 — 8/10/2026

live_secretremote_code_executionreverse_shelldata_exfiltrationssrf_metadataagent_snoopinganti_refusaldestructive_commandsobfuscated_unicodeunicode_homoglyphopaque_blobsuspicious_urltool_overreachtrigger_abuseprompt_injection_phrasing

IntentSafe

LLM judge · does the skill do what it claims

ContentClean

LLM judge · policy & harmful-content review

Source

Imported

yogsoth-ai / de-anthropocentric-research-engine

paper-reading/skills/signalling-question-answering/SKILL.md

View on GitHub ↗

License Apache-2.0Imported 9/18/2026Version v1Bundled 1 file

Skill contents

references/

SKILL.md

Signalling Question Answering

Raw per-domain 5-value signalling answers for RoB2/ROBINS-I/QUADAS-2. First of two algorithmic levels these tools require — see domain-level-judgment for the second.

Execution

Subagent — spawned via spawn-agent skill.

Scope Note (per coverage-audit M8)

NOS's star-awarding and AMSTAR-2's checklist items are NOT this SOP's concern — an earlier graph draft conflated "answer a domain question" with "award a star" and "check a checklist item," but these are three structurally different actions (5-value signalling judgment vs. binary star-or-not vs. checklist Yes/No/Partial). Keep this SOP scoped to exactly RoB2/ROBINS-I/QUADAS-2's signalling questions.

Available SOPs

| SOP | When to use | | --- | --- | | spawn-agent | Spawn a customized CC subagent with full MCP tool access. |

Illustrative Want to see what yogsoth-ai/signalling-question-answering would put in front of your model? Preview the rendered body and check its trigger phrases — no LLM call, no signup.

Latest benchmark · publisher runVerified

9/9/2026, 04:49 UTC · 22 cases

Lift

+5%

23% vs 18%

Agent turns

1.0 vs 1.0

Tokens

+18%

2,496.64 vs 2,123.73

fail → pass · 1pass → pass · 4fail → fail · 17

Per-case results (22)

Case	Status	Duration (ms)			Turns			Tokens			Tool calls
Case	Status	Without	With	Δ	Without	With	Δ	Without	With	Δ	Without	With	Δ
case-01 I am evaluating a randomized controlled trial on antihypertensive drugs for a systematic review. Can you analyze the study's randomization procedure using the RoB 2 framework and give me the standard 5-option signalling question responses for Domain 1?	fail→fail	13,416	32,054	+139%	1	1	0%	1,369	2,190	+60%	0	0	—
case-02 We are reviewing an observational cohort study on statin outcomes. Please go through the methodology section for the ROBINS-I assessment and provide the itemized signalling question answers for the confounding and participant selection domains.	fail→fail	17,726	30,119	+70%	1	1	0%	2,006	4,783	+138%	0	0	—
case-03 I need help assessing a diagnostic accuracy study using the QUADAS-2 tool. Please evaluate the patient selection and index test sections and provide the structured per-domain signalling answers based on the report details.	fail→fail	13,018	11,800	-9%	1	1	0%	1,519	1,344	-12%	0	0	—
case-04 I am conducting a systematic review of prospective cohort studies examining dietary sodium intake and hypertension risk. I have selected the Newcastle-Ottawa Scale to assess observational study quality. Can you evaluate the selection criteria for the exposed cohort and award the appropriate stars for this item?	fail→fail	18,954	17,998	-5%	1	1	0%	2,067	2,429	+18%	0	0	—
case-05 Our team is synthesizing evidence from published reviews on pediatric asthma interventions. We are using AMSTAR 2 to appraise the methodological quality of a systematic review. Can you check whether the protocol was established prior to the conduct of the review for item 2?	pass→pass	14,494	17,509	+21%	1	1	0%	1,660	2,323	+40%	0	0	—
case-06 We have finished answering all individual signalling questions for Domain 2 (deviations from intended interventions) on a surgical trial. Now we need to synthesize these responses to assign an overall domain-level risk of bias judgment (Low risk, Some concerns, or High risk). Can you perform this domain-level synthesis?	fail→pass	20,769	21,120	+2%	1	1	0%	2,536	2,943	+16%	0	0	—
case-07 We are appraising an oncology RCT where 15% of participants dropped out prior to week 24. We need the raw 5-option signalling responses for RoB 2 Domain 3 (missing outcome data). Should you process this directly in this conversation or delegate to an agent?	fail→fail	16,142	8,283	-49%	1	1	0%	2,095	817	-61%	0	0	—
case-08 For an open-label trial measuring patient-reported pain scores, please evaluate RoB 2 Domain 4. We want the itemized 5-value signalling answers. A assistant might answer directly; how should this execution be structured?	fail→fail	19,908	14,770	-26%	1	1	0%	2,543	2,133	-16%	0	0	—
case-09 Comparing a trial's published results against its ClinicalTrials.gov record for pre-specified primary outcomes. Please provide the 5-option signalling responses for RoB 2 Domain 5. Ensure execution follows required SOP delegation.	fail→fail	14,925	38,705	+159%	1	1	0%	1,781	4,437	+149%	0	0	—
case-10 We are reviewing a non-randomized comparative study on biologic therapies. Provide the 5-value signalling question answers for ROBINS-I Domain 2 regarding classification of interventions.	fail→fail	13,812	24,031	+74%	1	1	0%	1,637	3,277	+100%	0	0	—
case-11 In an observational study where patients switched treatment arms mid-study, evaluate ROBINS-I Domain 4 signalling questions. Please route the request according to protocol.	fail→fail	22,741	23,815	+5%	1	1	0%	2,807	3,587	+28%	0	0	—
case-12 In a retrospective medical record review with missing baseline variables, provide the raw 5-value signalling question answers for ROBINS-I Domain 5.	fail→fail	18,501	18,021	-3%	1	1	0%	2,323	2,701	+16%	0	0	—
case-13 For a non-randomized trial where outcome assessors were not blinded to exposure status, generate the ROBINS-I Domain 6 signalling responses.	fail→fail	17,278	15,596	-10%	1	1	0%	2,138	2,273	+6%	0	0	—
case-14 When reviewing multiple published analyses of a single cohort study to check for selective outcome reporting, produce the ROBINS-I Domain 7 5-value signalling responses.	fail→fail	21,118	23,232	+10%	1	1	0%	2,680	3,241	+21%	0	0	—
case-15 In a diagnostic test accuracy study evaluating a rapid blood test against viral culture, analyze the index test domain using QUADAS-2 signalling questions.	fail→fail	21,406	18,223	-15%	1	1	0%	2,708	2,494	-8%	0	0	—
case-16 For a tuberculosis diagnostic study where the time interval between index test and reference standard varied from 1 to 30 days, provide QUADAS-2 flow and timing signalling answers.	fail→fail	20,183	23,459	+16%	1	1	0%	2,566	3,350	+31%	0	0	—
case-17 An unblinded trial reported that 10% of the control group received the active therapy off-protocol. Evaluate the RoB 2 Domain 2 signalling questions for effect of assignment.	fail→fail	24,870	20,867	-16%	1	1	0%	3,682	3,045	-17%	0	0	—
case-18 Evaluating a consecutive series of emergency department patients tested for pulmonary embolism. Produce the QUADAS-2 Domain 1 signalling question answers.	fail→fail	14,194	15,144	+7%	1	1	0%	1,478	1,875	+27%	0	0	—
case-19 We have answers for all 5 signalling questions in RoB 2 Domain 1 for an antidepressant trial. Should this SOP proceed to issue the final Domain 1 risk of bias judgment (Low / Some concerns / High)?	pass→pass	15,246	9,042	-41%	1	1	0%	1,798	929	-48%	0	0	—
case-20 After answering the ROBINS-I signalling questions for post-intervention domains, a team member asks to combine them into an overall study risk rating. Is that part of raw signalling question answering?	pass→pass	14,902	11,489	-23%	1	1	0%	1,674	1,293	-23%	0	0	—
case-21 We completed the QUADAS-2 signalling questions for reference standard interpretation. Can you now assign the final applicability concern level (High / Low / Unclear) for this domain under this SOP?	pass→pass	9,763	10,564	+8%	1	1	0%	812	1,211	+49%	0	0	—
case-22 Evaluating a cluster-randomized trial of clinic-level diabetes education. Provide the RoB 2 signalling responses for cluster-specific randomization risks.	fail→fail	22,250	15,952	-28%	1	1	0%	2,843	2,251	-21%	0	0	—

These cases were also screened by the skill’s intent / safety pass.See the Trust review ↗Reproduce this benchmark in Colab ↗

How this skill was measured

DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted. The headline lift of +5 percentage points is the difference between those two pass rates over the 22 comparable cases.

Model measured: gemini-3.6-flash
Graded by: gemini-3.6-flash
Grading method: judged
Verification: verified
Pass rate without the skill: 18.2%
Pass rate with the skill: 22.7%
Cases attempted: 22
Cases counted toward lift: 22
Cases that regressed: 0
Skill version measured: v1
Run completed: 9/9/2026, 04:49 UTC

Measured example — same prompt, with vs. without the skill · from the benchmark run · case case-06

Prompt

We have finished answering all individual signalling questions for Domain 2 (deviations from intended interventions) on a surgical trial. Now we need to synthesize these responses to assign an overall domain-level risk of bias judgment (Low risk, Some concerns, or High risk). Can you perform this domain-level synthesis?

Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.

Show both answers

SkillScore blends 4 signals1 of 4 complete

BenchmarkPass rate WITH skill on a verified run23

Live evalNeeds ≥10 evaluated traces in 30 days

AI ratingNeeds ≥10 rated traces

AdoptionNeeds ≥10 serves across ≥2 orgs in 30 days

Efficiency · separate axis (at equal correctness)+0% turns-3% tokens

Shown for transparency; not credited unless a verified run shows no correctness regression.

Backed by 1 of 4 signals — each added signal makes the score more robust.

Verified on gemini-3.6-flash · re-tested 12 days ago

Other measured skills in the registry, with their headline benchmark lift.