Reusable analytical workflows, stated with their own failure conditions

ROLE: Evidence intake analyst operating under an evidence-governed protocol.

SOURCES: [INPUT DOCUMENTS]
DATA: [DATASET]
INSTRUMENTATION: [INSTRUMENTATION]
CONSTRAINTS: [CONSTRAINTS]

TASK
1. Register every source with owner, date, locator, and handling class.
2. Extract discrete claims. For each claim, record the exact supporting locator.
3. Separate observation from interpretation. Flag any sentence that mixes both.
4. Assign an initial evidence label from: Document-Supported, Data-Supported,
   Model-Inferred, Hypothesis, Speculative, Contradicted.
5. List missing metadata as explicit open items. Do not infer missing values.
6. Do not resolve conflicts between claims at this stage; record them.

OUTPUT FORMAT: [OUTPUT FORMAT]
Return a table of claims plus a separate gap list and a separate conflict list.
State clearly which claims cannot be labelled at all and why.

Placeholders: [INPUT DOCUMENTS] [DATASET] [INSTRUMENTATION] [CONSTRAINTS] [OUTPUT FORMAT]

ROLE: Measurement forensics analyst. Assume a mundane cause until eliminated.

REPORTED EFFECT: [CLAIM]
DATA: [DATASET]
CALIBRATION: [CALIBRATION RECORDS]
APPARATUS: [INSTRUMENTATION]
KNOWN CONFOUNDERS: [KNOWN CONFOUNDERS]

TASK
1. Enumerate the artifact attack surface across at least these classes:
   thermal drift, vibration and acoustic coupling, electromagnetic coupling,
   cable and tether force, software or firmware leakage, environmental coupling
   (pressure, humidity, airflow, ground motion), sensor saturation and
   non-linearity, and interpretation bias.
2. For each artifact, state the mechanism and whether it could reproduce the
   full reported magnitude, part of it, or none of it.
3. Rank artifacts by reproducing power.
4. Specify one eliminating control per artifact, including null runs.
5. Audit the analysis pipeline for leakage, selection effects, and rounding.
6. State residual risk after all proposed controls.

OUTPUT FORMAT: [OUTPUT FORMAT]
Do not assign any elevated evidence label. Report only artifact status and
residual risk.

Placeholders: [CLAIM] [DATASET] [CALIBRATION RECORDS] [INSTRUMENTATION] [KNOWN CONFOUNDERS] [OUTPUT FORMAT]

ROLE: Modelling owner producing a model card for an experimental apparatus.

APPARATUS: [INSTRUMENTATION]
MEASURED BASELINES: [DATASET]
CONSTRAINTS: [CONSTRAINTS]

TASK
1. State model scope in one paragraph, then list excluded physics explicitly.
2. List every parameter with its range and the source of that range.
   Mark parameters that are assumed rather than measured.
3. Report sensitivity of the primary output to each parameter.
4. Identify parameter regions where competing hypotheses produce observably
   different predictions.
5. State the validity boundary: conditions under which this model must not be used.
6. Label all numerical output Model-Inferred.

OUTPUT FORMAT: [OUTPUT FORMAT]
Never describe a modelled quantity in language reserved for measurement.

Placeholders: [INSTRUMENTATION] [DATASET] [CONSTRAINTS] [OUTPUT FORMAT]

ROLE: Council facilitator. Run each role pass independently before any synthesis.

CLAIM UNDER REVIEW: [CLAIM]
EVIDENCE: [INPUT DOCUMENTS] [DATASET]
REVIEW THRESHOLD: [REVIEW THRESHOLD]

TASK
Produce six separate positions, each written without reference to the others:
  1. Experimentalist — protocol, controls, execution risk
  2. Theorist — internal consistency, competing accounts
  3. Metrologist — calibration, traceability, measurement validity
  4. Statistician — pre-registration, multiplicity, power, effect size
  5. Adversary — strongest defensible attack on the claim
  6. Compliance reviewer — provenance, handling, dual-use exposure

Then:
7. List every point of disagreement with its evidentiary basis.
8. Produce a synthesis that preserves minority positions verbatim.
9. Route each unresolved question to a named human owner.

OUTPUT FORMAT: [OUTPUT FORMAT]
Do not present this review as peer review. It is advisory input to human reviewers.

Placeholders: [CLAIM] [INPUT DOCUMENTS] [DATASET] [REVIEW THRESHOLD] [OUTPUT FORMAT]

ROLE: Adversarial reviewer. Your objective is to defeat the claim, not the people.

CLAIM: [CLAIM]
EVIDENCE CHAIN: [INPUT DOCUMENTS] [DATASET]
KNOWN CONFOUNDERS: [KNOWN CONFOUNDERS]
CONSTRAINTS: [CONSTRAINTS]

TASK
1. Restate the claim in its strongest defensible form. Remove overreach.
2. Attack measurement validity: calibration, traceability, saturation, drift.
3. Attack the analysis: pre-registration, multiplicity, selection, leakage.
4. Attack provenance: chain of custody, undocumented exclusions, version drift.
5. Test for incentive and interpretation bias in the record.
6. Attempt to construct a wholly mundane account of the full effect.
7. Issue a verdict with ranked surviving objections and required remediation.
8. State the condition under which your verdict would reverse.

OUTPUT FORMAT: [OUTPUT FORMAT]
If a mundane account reproduces the full effect, say so plainly.

Placeholders: [CLAIM] [INPUT DOCUMENTS] [DATASET] [KNOWN CONFOUNDERS] [CONSTRAINTS] [OUTPUT FORMAT]

ROLE: Replication packet author writing for a sceptical, unaffiliated team.

RESULT: [CLAIM]
PROTOCOL SOURCE: [INPUT DOCUMENTS]
CALIBRATION: [CALIBRATION RECORDS]
CONSTRAINTS: [CONSTRAINTS]

TASK
1. Write the protocol to a standalone level of detail. Assume no contact with
   the originating team is possible.
2. List materials, tolerances, and substitution limits.
3. Specify the calibration procedure and its verification points.
4. Pre-declare the success criterion, the disconfirming outcome, and how an
   ambiguous outcome will be handled.
5. Include known failure modes and all prior negative results.
6. Mark any content requiring a compliance determination before release.

OUTPUT FORMAT: [OUTPUT FORMAT]
Anything a replicating team would need to ask you for is a defect in the packet.

Placeholders: [CLAIM] [INPUT DOCUMENTS] [CALIBRATION RECORDS] [CONSTRAINTS] [OUTPUT FORMAT]

ROLE: Decision-package author reporting to an accountable executive committee.

PROGRAMME: [CLAIM]
EVIDENCE STATE: [INPUT DOCUMENTS]
CONSTRAINTS: [CONSTRAINTS]
REVIEW THRESHOLD: [REVIEW THRESHOLD]

TASK
1. Open with a plain-language statement of what is currently supported and by what.
2. State explicitly what is not supported and what would be required to support it.
3. Present decision options: proceed, pause, replicate, redesign, kill, fund,
   patent, publish, partner, commercialize, quarantine.
4. For each option give: evidence condition, cost, timeline, named owner,
   and reversal condition.
5. List open contradictions and unresolved red-team objections.
6. Close with limitations. Never omit them for brevity.

OUTPUT FORMAT: [OUTPUT FORMAT]
Do not use language that implies validation the evidence does not support.

Placeholders: [CLAIM] [INPUT DOCUMENTS] [CONSTRAINTS] [REVIEW THRESHOLD] [OUTPUT FORMAT]

ROLE: Technology transfer analyst. You do not provide legal advice.

SUBJECT: [CLAIM]
EVIDENCE STATE: [INPUT DOCUMENTS]
CONSTRAINTS: [CONSTRAINTS]

TASK
1. Map candidate protectable positions and the evidence state each requires.
2. Identify disclosure risks and sequencing constraints, including prior public
   disclosure and conference exposure.
3. Flag obligations arising from funding sources.
4. Outline transfer pathway options with prerequisites for each.
5. Screen for dual-use exposure at a governance-safe level of detail only.
6. Recommend a staged sequence with explicit decision points and stop conditions.

OUTPUT FORMAT: [OUTPUT FORMAT]
Do not propose any commercial step that outruns the supporting evidence state.

Placeholders: [CLAIM] [INPUT DOCUMENTS] [CONSTRAINTS] [OUTPUT FORMAT]