investigation

Name: investigation
Availability: InStock
Author: dylan-conlin

Coordinate agent experiments with enforced tests and evidence-backed conclusions.

Updated Aug 23, 2026

Search Tags:#workflow #testing #investigation #evidence #error-recovery #knowledge-management #probes

Ready to run

npx skills add https://github.com/dylan-conlin/orch-go --skill investigation-dylan-conlin

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: investigation
Source: https://github.com/dylan-conlin/orch-go/tree/main/skills/src/worker/investigation
Command: npx skills add https://github.com/dylan-conlin/orch-go --skill investigation-dylan-conlin

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires worker-base.

What problem does it solve?

Records and validates agent work by enforcing tests and structured evidence, turning raw outputs into trustworthy knowledge.

Core Features & Use Cases

Probes and investigations discipline: choose probe mode when injected model claims exist and run investigations otherwise; maintain D.E.K.N. summaries, support prior-work relationships, and enable error recovery.

Quick Start

Create a new investigation and immediately checkpoint your question and initial finding.

Frequently Asked Questions about investigation

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I document agent experiments and enforce evidence-backed conclusions?▼

You document agent experiments by enforcing real testing and preserving evidence, turning raw agent outputs into structured, trustworthy knowledge artifacts for recovery and self-review.

What is the best way to validate injected model claims during testing?▼

Validating injected model claims uses probe-mode, which applies targeted testing to verify the claims and produces structured artifacts that preserve the evidence for later review.

How do I recover from errors during an agent investigation workflow?▼

Error recovery during an agent investigation is supported by maintaining D.E.K.N. summaries and prior-work relationships, allowing you to self-review, checkpoint, and resume structured experiments.

Can I run full investigations without injected model claims?▼

Yes, you can run full investigations when no injected model claims exist, coordinating comprehensive agent experiments that enforce tests and produce structured, evidence-backed conclusions.

Does the investigation workflow require a specific base setup?▼

Yes, the investigation workflow requires the worker-base dependency to provide the foundational environment needed for coordinating probes, enforcing tests, and generating structured artifacts.

Why should I use structured testing for agent knowledge management?▼

Structured testing for agent knowledge management ensures raw outputs are validated against real evidence, preventing unverified conclusions and producing trusted, recoverable knowledge artifacts.