AI agent skill · by @yogsoth-ai

Define Evidence Protocol

Define research questions, evidence scope, queries/concepts, inclusion/exclusion rules, provenance requirements, and stopping evidence for an acquisition task.

Archived 2026-10-23 · Source updated 2026-09-29 · License: Apache-2.0. Check the original repository for the latest version.

SKILL.md instructions

# define-evidence-protocol

## Purpose

Define research questions, eligible evidence scope, query concepts, inclusion/exclusion rules, provenance requirements, and stopping evidence.

## Input contract

```yaml
required: [research_question, evidence_scope, inclusion_rules]
optional: [exclusion_rules, query_concepts, quality_rubric, stopping_policy]
constraints: [eligible universe, provenance fields, and stopping evidence must be explicit]
```

## Procedure

1. State the question, target population or domain, comparison, outcomes, and time boundaries.
2. Define eligible evidence, exclusion rules, source-independence requirements, and extraction fields.
3. Specify batch boundaries, novelty dimensions, and relative stopping evidence.
4. Record unresolved scope choices before acquisition begins.

If the protocol fixes inclusion, exclusion, and stopping rules, consider `select-seed-evidence` as the next tactic.

## Output contract

```yaml
produces: [evidence_protocol, eligible_universe, query_concepts, screening_rules, provenance_requirements, stopping_evidence]
delta_fields: [decisions, assumption_updates, uncertainties, open_questions]
```

## Quality gates

- The eligible universe and exclusion rules are operational.
- Stopping evidence names numerator, denominator, batch increment, source references, direction, threshold, and rationale when relative quantities apply.

## Failure and counterexamples

Do not use provider names as a protocol, and do not declare coverage from search volume without a defined eligible universe.

## Provenance map

- `resolved: define-search-protocol`
- `resolved: inclusion-criteria-design`