| name | development-data-caution |
| description | Responsible-language and claim-discipline skill for development research and operations. Use when describing vulnerable households, survey staff, payment providers, field quality flags, welfare changes, targeting results, or policy/operational implications from synthetic or real World Bank-style data.
|
| workflow_stage | communication |
| version | 1.0.0 |
| tags | ["development","responsible-language","caveats","operational-risk"] |
Development Data Caution
Use this skill whenever writing findings, slide claims, QA reports, or recommendations.
Claims
- Say "descriptive change" unless an identification strategy supports causal language.
- Separate targeting, payment delivery, welfare, and data quality findings.
- State data limitations next to operationally sensitive claims.
- Prefer "suggests", "flags", and "requires review" over definitive explanations when evidence is incomplete.
People And Partners
Use neutral language for households, enumerators, providers, and counterparts.
Good:
- "households flagged for follow-up";
- "provider records require reconciliation";
- "enumerator-level pattern needs supervisor review";
- "district has lower observed completion."
Avoid:
- stigmatizing households;
- implying fraud or misconduct without confirmation;
- blaming local staff from data flags alone.
Sensitive Outputs
Before producing or sharing operational lists:
- check whether records contain PII;
- minimize unnecessary fields;
- state that synthetic Lumora data are for training only;
- recommend human review before action.
Use:
