Skip to content

feat: add meta-analyst agent to mft-research-experts#39

Open
JulyMarc wants to merge 3 commits intoDeevsDeevs:mainfrom
JulyMarc:feature/meta-analyst
Open

feat: add meta-analyst agent to mft-research-experts#39
JulyMarc wants to merge 3 commits intoDeevsDeevs:mainfrom
JulyMarc:feature/meta-analyst

Conversation

@JulyMarc
Copy link

@JulyMarc JulyMarc commented Mar 3, 2026

Summary

  • Adds meta-analyst agent to the mft-research-experts group
  • Domain-agnostic data analysis expert that conducts multi-lens analysis: exploratory, comparative, temporal, perspective
  • Designed for analytical reports, market monitoring, operational metrics, and any dataset that needs expert interpretation

Agent capabilities

  • Domain immersion before analysis (reads context, understands metrics semantics)
  • Multi-lens analysis framework (exploratory → comparative → temporal → perspective)
  • Structured output with confidence levels and actionable recommendations

🤖 Generated with Claude Code

JulyMarc and others added 3 commits March 3, 2026 13:36
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add protections against common LLM data errors:
- Column provenance tracking after joins
- Temporal integrity checks (lookahead prevention)
- Aggregation validity rules
- Join semantics verification
- Mandatory post-transformation checks

Covers 6 error classes: lookahead bias, wrong aggregation,
join direction, off-by-one temporal, column confusion,
survivorship bias.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
@yk4r2
Copy link
Collaborator

yk4r2 commented Mar 9, 2026

Overlaps strongly with data-sentinel, causal-detective, mft-strategist; maybe it would be better to complete any of these?

Data integrity + causal validation might be delegated (so should be rewritten as the escalation + delegation triggers)

Needs a narrower contract, I think validation gatekeeping + research orchestration might be redundant.

The (probable) best new value is to my mind domain immersion + arbitrary datasets' analysis.

Workflow should end with "escalate to <specialist's name>", not with "run the full investigation pipeline myself"

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants