What is dq-hypothesizer?

A data-quality hypothesis writer for a dbt model or data-lineage node. dbt is a tool that builds documented data models; each hypothesis is written as a claim that one read-only SQL query can confirm or disprove.

Use it to examine changed or investigated data models, use available business and engineering findings, and produce runnable checks with clear interpretations of their results.

It turns vague concerns about incorrect data into testable questions before a separate checker evaluates them. This helps keep observations, suspicions, and verdicts distinct.

dq-hypothesizer — Agent for Claude Code by kbichave/skills

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

Clone the repo

git clone --depth 1 https://github.com/kbichave/skills

Made for: Claude Code.

Or install deep, the plugin that ships this one along with the rest of its 5 skills, 22 agents, 6 hooks.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

README.md

[![agentmods](https://agentmods.dev/badge/agents/kbichave/skills/dq-hypothesizer/github.svg)](https://agentmods.dev/agents/kbichave/skills/dq-hypothesizer)

Your own site

<a href="https://agentmods.dev/agents/kbichave/skills/dq-hypothesizer"><img src="https://agentmods.dev/badge/agents/kbichave/skills/dq-hypothesizer/github.svg" alt="Measured on agentmods" height="20"></a>

Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.

Your own site · 80×15

<a href="https://agentmods.dev/agents/kbichave/skills/dq-hypothesizer"><img src="https://agentmods.dev/badge/agents/kbichave/skills/dq-hypothesizer.svg" alt="Reviewed on agentmods" width="80" height="20"></a>

Per session 142 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.

When invoked 2,384 The whole file, excluding the scripts and references it only reads on demand.

Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.

Origin original No closer match found in the catalogue.

Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

Model	Per session	Once invoked
Fable 5.1	$0.00142	$0.02384
Opus 5.5	$0.00057	$0.00954
Sonnet 5.5	$0.00028	$0.00477
Haiku 4.5	$0.00014	$0.00238

Measured 11d ago against content hash 1ed3faf3a226, method: parsed. Prices are Anthropic first-party input rates as of 2026-10-07, from the pricing page.

Security

Grade A, and why

dq-hypothesizer scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 11d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

agents/dq-hypothesizer.md · 167 lines

How it starts

The opening of the file, as written. The whole thing — 167 lines — stays where its author put it; the contents beside it link to each section on GitHub.

DQ Hypothesizer (data tier: `dq-hypothesizer`)

Work out what could be silently wrong with one node, and write each suspicion as a claim a single query can prove or disprove. You form hypotheses; dq-prover tests them. Facts you observe while gathering evidence go in evidence_log, never in a verdict.

Resolve tool roles (query, describe, model_health, lineage) through references/dq/warehouse-tools.md before the first call, and follow its read-only and personal-data rules on every query.

Input

The hand-off in references/dq/scope.md, plus:

diff_hunk: the changed lines of the model, or none (investigate runs).
data_eng_findings: the panel's DE-* findings on this node, each with anchor, hypothesis and disproof_sql where the expert wrote one.
model_context and domain_context: output of model-context and domain-context-reader for the scope, or "none".
profile: the deployment profile JSON (environments, ci_schema_pattern, pii_column_patterns), or null.
budget: light (default) or full.

Never fetch what the input already holds. If no context is given, get the node's relation, columns, and health yourself and say the context was missing.

In the review panel

The orchestrator also runs the standard checks (references/dq/standard-hypotheses.md, H1 to H4, H8, H10, H11 per relation; H9 per edge; fan-out per changed join) and every disproof_sql through dq-prover separately. So:

write only the hypotheses the evidence points to beyond those checks;
the diff hunk is the first lead, then the data-eng findings;
a data-eng finding that already carries a disproof_sql gets no duplicate hypothesis; extend it only with a competing cause or a dependent effect;
every test_sql stays runnable on its own, because provers run them in batches.

Workflow

- [ ] 0. Read references/dq/lenses.md
- [ ] 1. Read the input; list what is known, what is not, and the leads
- [ ] 2. Read the node's SQL and config; mark logic worth suspecting
- [ ] 3. Pull only the extra evidence the lenses need (references/dq/evidence.md)
- [ ] 4. Run the allowed previews to see the data's shape
- [ ] 5. Go through every lens; write hypotheses or "n/a: why"
- [ ] 6. Return the JSON

Read the full file on GitHub · 167 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

11d ago First seen · 167 lines · 142 tokens per session scan A 1ed3faf3a226

Subscribe to this mod's changes

dq-hypothesizer is an agent published in the GitHub repository kbichave/skills (2 stars, last pushed 12d ago), licensed MIT. It adds 142 tokens to every session and 2,384 once invoked, about $0.0006 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-10-08.

Analytics and metrics ai-planning claude-code-plugin implementation planning multi-llm

corporate-strategy

Corporate Strategy (CSO). Owns plugins/corporate-strategy/ and nothing else. Delegate work in this department's remit here.

cbrock84/headcount · 33 tokens

demand-generation

Demand Generation (CMO). Owns plugins/demand-generation/ and nothing else. Delegate work in this department's remit here.

cbrock84/headcount · 30 tokens

canvas-create-screens

Groups approved functional intent into task-oriented screens, in-screen states and navigation without choosing Canvas controls.

microsoft/power-platform-skills · 24 tokens

debugger

Diagnoses and fixes failed modules using root-cause analysis, not guessing.

TT-Wang/forge · 17 tokens

scope

IP strategy — trademark clearance, patent landscape, open source license compliance, IP assignment.

tonone-ai/tonone · 18 tokens

motion-critic

Use when: a motion-expert draft or master video needs a blind, frames-only quality verdict before an owner gate or a done claim. Do NOT use for: building or fixing a film (motion-expert), code review of render scripts (sniper), or verifying claims about the pipeline (challenger).

fusengine/agents · 66 tokens