data-aggregator
Skill1 starsChanged 5 months agoApache-2.0View on GitHub
Data Aggregator — performs secondary processing on raw data from a data executor: validation, cleaning, aggregation, year-over-year/month-over-month comparison, statistical enhancement and annotation. 4-Phase forced orchestration via Phase-Orchestrator. Depends on Data-Analyst component. Triggers: aggregation, statistics, grouping, YoY/MoM, TOP ranking, data summarization.
In this repository4 other files it gives its agents.
- evidence-chainSkill
- nl2-querySkill
- scoring-engineSkill
- visualization-rendererSkill
The file
We split this skill into the rules it gives an agent
1 rule. About 0 are common, 1 is distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 130
Conditional execution: Only when comparison_config is not null.
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Data Aggregator
Core data aggregation component. Depends on Data-Analyst, orchestrated via Phase-Orchestrator.
Parameters
| Parameter | Type | Required | Default | Description |
| raw_data | JSON array | Yes | - | Raw data from data executor |
| aggregation_config | JSON | Yes | - | Aggregation configuration (operation types and parameters) |
| comparison_config | JSON | No | null | YoY/MoM comparison configuration |
| null_strategy | Option | No | ignore | Null strategy: ignore/fill_zero/mark |
Supported Aggregation Operations
| Operation | Description | Input Requirement | Output |
| count | Count | Any field | Number |
| sum | Sum | Numeric field | Number |
| avg | Average | Numeric field | Number |
| max | Maximum | Numeric/Date field | Number/Date |
| min | Minimum | Numeric/Date field | Number/Date |
| percentile | Percentile | Numeric field + p value | Number |
| group_by | Group by | Any field + aggregate function | Grouped results |
| top_n | Top N | Sort field + N value | Sorted results |
| compare | YoY/MoM | Numeric field + time dimension | Comparison results |
| derive | Derived calc | Existing fields + formula | New field |
4-Phase Forced Orchestration Protocol
Orchestrated via Phase-Orchestrator. Each Phase loads components via skill tool and launches a general-type sub-Agent via task tool.
Phase 1: Data Validation and Cleaning
Load component: skill -> data-analyst
Execution: task tool launches general-type sub-Agent, prompt:
You are the Phase 1 executor of the Data Aggregator. Validate and clean the raw data.
Input: raw_data (JSON array), null_strategy (ignore/fill_zero/mark)
Step 1: Format validation
- Check raw_data is a valid JSON array
- Check each record has consistent structure (same field names)
- Check numeric fields are number type (not string)
Step 2: Null handling (based on null_strategy)
- ignore -> skip records with null values
- fill_zero -> replace null values with 0
- mark -> keep null values but flag as "data_missing"
Step 3: Outlier detection
- Numeric fields: detect values beyond 3-sigma range
- Date fields: detect future dates or very old dates
- Enum fields: detect values not in known enum list
- Flag but do not remove outliers
Step 4: Data summary
- Total rows, field count
- Each field's type, null rate, unique count
- Numeric fields' min/max/avg/median
Output JSON:
{
"validated_data": [...],
"data_quality": {
"total_rows": 0, "total_fields": 0,
"null_count": 0, "null_rate": 0.0,
"outlier_count": 0, "outlier_fields": []
},
"field_summary": {
"<field_name>": {"type": "number|string|date", "null_rate": 0.0, "unique_count": 0, "min": null, "max": null, "avg": null}
},
"cleaning_actions": ["Replaced null with 0 (3 records)", "Flagged 2 outliers"],
"phase_status": "success|warning"
}Phase 2: Aggregation Calculation
Load component: skill -> data-analyst
Execution: task tool launches general-type sub-Agent, prompt:
You are the Phase 2 executor of the Data Aggregator. Execute aggregation calculations.
Input: Phase 1 output validated_data, aggregation_config
Execute based on aggregation_config:
group_by: group by specified field, apply aggregate function per group
top_n: sort by specified field, take top N, add rank and percentile columns
percentile: calculate specified percentile values (supports 25th/50th/75th/90th/95th/99th)
derive (derived calculation): calculate new fields from existing ones
- ratio: ratio
- difference: difference
- growth_rate: growth_rate = (current - previous) / previous * 100
Output JSON:
{
"aggregated_data": [...],
"aggregation_meta": {
"operations_performed": ["group_by:plan_type", "count", "avg:amount"],
"groups_count": 0,
"result_rows": 0
},
"phase_status": "success"
}Phase 3: Comparison Calculation (YoY/MoM)
Conditional execution: Only when comparison_config is not null; otherwise skip and pass Phase 2 output directly to Phase 4.
Execution: task tool launches general-type sub-Agent, prompt:
You are the Phase 3 executor of the Data Aggregator. Calculate year-over-year and month-over-month comparisons.
Input: Phase 2 output aggregated_data, comparison_config
YoY calculation: same dimension, current period vs same period last year
YoY = (current - last_year) / last_year * 100%
Annotation: up / down / flat
MoM calculation: same dimension, current period vs previous period
MoM = (current - last_month) / last_month * 100%
Annotation: up / down / flat
Notes:
- Same period last year not available -> annotate "no_comparable_data"
- Base is 0 -> annotate "base_zero_cannot_calculate"
- Change exceeds 100% -> annotate "large_fluctuation"
Output JSON:
{
"comparison_data": [
{
"dimension": "Premium Plan", "current": 47, "previous": 35, "last_year": 28,
"mom_change": 34.3, "mom_direction": "up",
"yoy_change": 67.9, "yoy_direction": "up"
}
],
"comparison_meta": {
"current_period": "2026-05", "previous_period": "2026-04",
"yoy_period": "2025-05", "data_completeness": "full|partial"
},
"phase_status": "success|partial"
}Phase 4: Statistical Enhancement and Annotation
Execution: task tool launches general-type sub-Agent, prompt:
You are the Phase 4 executor of the Data Aggregator. Add statistical enhancement and smart annotations to aggregation results.
Input: Phase 3 output comparison_data (or Phase 2 aggregated_data if Phase 3 skipped)
Step 1: Derived metrics
- Proportion: each dimension value / total * 100%
- Cumulative: cumulative value and cumulative percentage by sort order
- Z-Score: deviation of each dimension from the mean
Step 2: Smart annotation
- YoY/MoM change exceeds 20% -> focus_attention
- Z-Score > 2 -> anomaly
- Rank change exceeds 3 positions -> rank_change annotation
- CR3 proportion > 70% -> high_concentration
Step 3: Summary generation
- One-sentence summary of data characteristics
- List TOP3 attention items
- Flag anomalies requiring special attention
Output JSON:
{
"enhanced_data": [...],
"annotations": [
{"dimension": "Premium Plan", "type": "significant_increase", "detail": "MoM growth 34.3%, well above average"}
],
"derived_metrics": {
"concentration_cr3": 0.72,
"gini_coefficient": 0.38,
"total": 140,
"average": 28
},
"summary": {
"one_liner": "Support tickets concentrated on Premium Plan (33.6%) and Standard Plan (27.1%)",
"top3_attention": ["Premium Plan: MoM growth 34.3%", "Standard Plan: proportion 27.1%", "Basic Plan: YoY decline 12%"],
"anomalies": ["Enterprise Plan ticket volume significantly below average"]
},
"phase_status": "success"
}Phase Data Passing
- Phase 1 -> Phase 2:
{ "validated_data": [...], "data_quality": {...}, "field_summary": {...} } - Phase 2 -> Phase 3:
{ "aggregated_data": [...], "aggregation_meta": {...} } - Phase 3 -> Phase 4:
{ "comparison_data": [...], "comparison_meta": {...}, "aggregated_data": [...] }
Degradation Strategy
| Condition | Degradation Behavior |
| Phase 1 data quality poor (null rate > 30%) | Degrade to descriptive statistics only |
| Phase 2 aggregation fails | Return raw data + error explanation |
| Phase 3 no historical data | Skip comparison, proceed directly to Phase 4 |
| Phase 4 annotation fails | Return aggregated data only, without annotations |
Usage Example
Input:
{
"raw_data": [
{"plan_type": "Premium Plan", "ticket_count": 47},
{"plan_type": "Standard Plan", "ticket_count": 38},
{"plan_type": "Basic Plan", "ticket_count": 25},
{"plan_type": "Lite Plan", "ticket_count": 18},
{"plan_type": "Enterprise Plan", "ticket_count": 12}
],
"aggregation_config": {
"operations": ["group_by", "top_n", "derive"],
"group_by": {"field": "plan_type"},
"top_n": {"field": "ticket_count", "n": 5, "order": "desc"},
"derive": [{"name": "percentage", "formula": "value / total * 100"}]
},
"comparison_config": {
"current_period": "2026-05",
"previous_period": "2026-04",
"yoy_period": "2025-05",
"metric": "ticket_count"
}
}Final output:
{
"enhanced_data": [
{"plan_type": "Premium Plan", "ticket_count": 47, "percentage": 33.6, "rank": 1, "mom": "up 34.3%", "yoy": "up 67.9%"},
{"plan_type": "Standard Plan", "ticket_count": 38, "percentage": 27.1, "rank": 2, "mom": "up 5.6%", "yoy": "up 18.8%"},
{"plan_type": "Basic Plan", "ticket_count": 25, "percentage": 17.9, "rank": 3, "mom": "down 12.0%", "yoy": "flat"},
{"plan_type": "Lite Plan", "ticket_count": 18, "percentage": 12.9, "rank": 4, "mom": "up 10.0%", "yoy": "up 28.6%"},
{"plan_type": "Enterprise Plan", "ticket_count": 12, "percentage": 8.6, "rank": 5, "mom": "flat", "yoy": "down 7.7%"}
],
"annotations": [
{"dimension": "Premium Plan", "type": "significant_increase", "detail": "MoM growth 34.3%, YoY surge 67.9%"}
],
"summary": {
"one_liner": "Support tickets concentrated on Premium Plan (33.6%) and Standard Plan (27.1%)",
"top3_attention": ["Premium Plan MoM +34%", "Standard Plan proportion 27%", "Basic Plan YoY decline 12%"]
}
}Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

