AgentLogistics

AgentLogistics is an open-source collection of AI skills for commercial logistics work. It covers warehousing, inventory, fulfillment, transportation, material handling, logistics systems, labor planning, continuous improvement, and selected source-sensitive specializations.

The repository gives general-purpose AI agents bounded procedures, evidence requirements, calculation checks, output contracts, and explicit safety and approval boundaries. It provides decision support; it is not a substitute for qualified professional, legal, regulatory, engineering, or operational review.

Current Status

Item	State
Public release	v0.1.0-public-preview
Roadmap coverage	AL-00 through AL-25 complete
v1 audit verdict	V1_PARTIALLY_READY
Local validation	Passing
Stable v1 release	Not released

The public preview is suitable for inspection, skill evaluation, and early workflow use. It is not a v1 release. The remaining v1 gates are live model evaluation, recurring source-freshness review, broader calculation fixtures, and final release-setting decisions. See the v1 release-candidate audit.

Who It Is For

• AI agent builders who need reusable logistics skills and role compositions.
• Warehouse, inventory, transportation, and distribution teams evaluating AI support workflows.
• Operations analysts who need structured prompts for investigations, calculations, planning, and improvement work.
• Contributors who want to add source-backed logistics knowledge in small, testable packages.

What It Can Do

The current public preview includes:

• 143 validated core skill packages;
• 14 validated professional skillsets;
• 42 validated specialization packages;
• 169 taxonomy candidates with duplicate-slug checks;
• shared formulas, glossaries, schemas, and output templates;
• deterministic scenarios, fixtures, routing expectations, and evaluation reports;
• Canada, United States, food cold-chain, dangerous-goods, and international logistics source maps;
• GitHub Actions validation and scheduled source-link auditing.

Core coverage includes receiving, storage, putaway, replenishment, picking, packing, shipping, returns, inventory control, capacity, slotting, freight, systems and data, warehouse labor, KPIs, bottleneck analysis, root cause analysis, and improvement planning.

Quick Start

Clone the repository and run the deterministic validation gate:

git clone https://github.com/jeremylongworth-source/AgentLogistics.git
cd AgentLogistics
.\scripts\validate-all.ps1

Prerequisites are Git, PowerShell, and Python available as python. The local repository checks do not currently require a package installation.

Expected result:

All AgentLogistics validation checks passed.

How To Use The Skills

Skills use progressive disclosure. Load the smallest package first, then load references only when the task requires them.

1. Choose an atomic skill, skillset, or specialization.
2. Read its SKILL.md.
3. Load referenced files as needed.
4. Gather required inputs and evidence.
5. Follow the procedure and output contract.
6. State assumptions, missing evidence, calculations, and review boundaries.

Start with the reference calculation skill:

skills/inventory-control/calculate-reorder-point/

It demonstrates the expected separation of procedure, references, agent metadata, calculations, validation, and tests. Other useful entry points include:

• investigate-inventory-discrepancy for conflicting receiving, system, count, pick, and adjustment evidence;
• plan-picking-wave for order-profile, cutoff, capacity, equipment, and labor constraints;
• audit-freight-charge for rate basis, accessorial, invoice, and shipment evidence review;
• triage-dangerous-goods-incident-logistics for source-bounded logistics triage and escalation support.

GitHub Skill Install

AgentLogistics can be previewed and installed through GitHub CLI agent skills. The public preview is namespaced by domain and should be pinned when reproducibility matters:

gh skill preview jeremylongworth-source/AgentLogistics inventory-control/calculate-reorder-point
gh skill install jeremylongworth-source/AgentLogistics inventory-control/calculate-reorder-point --agent github-copilot --scope project --pin v0.1.0-public-preview

See GitHub Copilot and gh skill setup for preview, install, pinning, verification, and maintainer publishing checks. See the public preview release notes for the release boundary.

Package Model

AgentLogistics uses three layers:

• Skills are atomic packages for one bounded logistics task.
• Skillsets compose existing atomic skills into professional roles without duplicating their procedures.
• Specializations isolate jurisdiction-specific, industry-specific, mode-specific, or otherwise source-sensitive material.

Universal core skills must not depend on specializations. Regulated and safety-sensitive work must preserve jurisdiction, authority, source, date, applicability, assumptions, evidence gaps, and qualified-review boundaries.

Repository Layout

AgentLogistics/
|-- .github/             GitHub workflows and contribution templates
|-- docs/                Architecture, standards, setup, audits, and releases
|-- scripts/             Local validation and source-link audit scripts
|-- shared/              Formulas, glossaries, schemas, and templates
|-- skills/              Atomic logistics skill packages
|-- skillsets/           Professional role compositions
|-- specializations/     Jurisdiction, industry, mode, and source-sensitive work
|-- tests/               Scenarios, fixtures, routing, and evaluation reports
|-- AGENTS.md            Repository instructions
|-- ROADMAP.md           Development authority
|-- CONTRIBUTING.md      Contribution rules and review expectations
|-- SECURITY.md          Security and safety reporting guidance
`-- README.md            Public entry point

Use the GitHub Wiki for guided orientation. The repository remains the source of truth for skills, standards, tests, validators, roadmap handoffs, and release artifacts.

Validation And Testing

Run the full local gate before contributing:

.\scripts\validate-all.ps1

The wrapper checks documentation, taxonomy, skill packages, skillsets, specializations, tests, shared foundations, and source metadata. The test model uses:

• scenario routing in tests/scenarios/ and tests/expected-routing.yaml;
• deterministic calculation and workflow fixtures in tests/fixtures/;
• evaluation reports in tests/evaluations/;
• GitHub Actions validation on pushes, pull requests, and manual dispatch;
• a scheduled and manual source-link audit for specialization source maps.

See Validation and testing, the evaluation standard, and the v1 hardening CI note for detailed contracts and current limitations.

Skillsets

Professional compositions currently include:

• warehouse operator;
• receiving specialist;
• inventory control specialist;
• warehouse supervisor and warehouse manager;
• logistics coordinator and transportation coordinator;
• warehouse planner and distribution manager;
• logistics systems analyst;
• continuous improvement specialist;
• logistics operations manager;
• fulfillment optimizer;
• material handling analyst.

See skillsets/README.md for the composition index and role boundaries.

Specializations

Source-sensitive material is isolated under specializations/:

• Canada
• United States
• Food and cold-chain logistics
• Dangerous goods
• International logistics

These packages support research, planning, evidence handling, and escalation. They do not approve compliance, classify goods for release, clear customs, release food, certify equipment or operators, or replace the applicable authority or qualified reviewer.

Limitations

AgentLogistics is decision-support content. Regulatory and safety-sensitive material is jurisdiction-specific unless a source clearly states otherwise.

AgentLogistics does not operate live WMS, TMS, ERP, OMS, YMS, LMS, WCS, WES, freight, customs, carrier, government, or safety systems.

It does not approve inventory adjustments, financial postings, carrier awards, rate acceptance, claims, supplier or customer commitments, live system edits, labor commitments, staffing decisions, equipment purchases, facility changes, structural rack changes, or building expansion.

It must not present output as legal advice, engineering approval, equipment certification, operator certification, regulatory approval, customs entry approval, food safety release, dangerous-goods classification approval, or professional compliance signoff. Verify current requirements with the applicable authority before acting on regulated or safety-sensitive material.

Development Status

Roadmap waves AL-00 through AL-25 are complete. The AL-25 audit recorded V1_PARTIALLY_READY, so the project remains pre-v1. The next development wave is v1 hardening:

Completion records:

AGENTLOGISTICS_AL_24_PUBLIC_READINESS_READY
AGENTLOGISTICS_AL_25_V1_RC_AUDIT_COMPLETE

• score representative skills, skillsets, specializations, and integration scenarios with live model evaluations;
• review recurring source-link results and record freshness decisions;
• expand independent numerical fixtures for calculation-heavy skills;
• confirm CI and repository-setting requirements;
• prepare v1 release notes and a tag only after the gates pass.

The development authority is ROADMAP.md. The evidence for the current decision is the AL-25 audit and its final handoff.

Contributing

Read CONTRIBUTING.md before opening an issue or pull request. Contributions should keep skills atomic, use authoritative sources, separate universal and jurisdiction-specific material, show calculation math, preserve safety boundaries, and add scenarios, fixtures, evaluations, or validators when behavior changes.

Run .\scripts\validate-all.ps1 before submitting.

Security And Safety

Use SECURITY.md for vulnerability and safety-sensitive content reporting. Do not publish secrets, credentials, customer data, shipment data, facility layouts, personnel records, or other sensitive operational details in public issues or pull requests.

License

AgentLogistics is licensed under the MIT License.