Move from a symptom to traceable evidence, a sourced decision, and a handover the next engineer can follow. SkillVault combines reusable AI workflows with field documents, worked examples, training, and an offline evidence logger.
The final package was checked beyond whether the files open. The skills, application, documents, workbooks, course, privacy boundary, and manifest were verified together.
Zero package-validation errors in the qualified buyer archive.
Purpose, triggers, safety gates, methods, examples, mistakes, and adaptation are present.
Source-gated judgment, calculations, exports, and recorded-not-judged behavior are covered.
The delivery package carries a SHA-256 manifest for integrity checking.
A Skill is not a prompt you paste in and hope for the best. It is a reusable workflow that tells a supported AI assistant when to engage, which authority boundaries apply, what evidence is missing, and what a complete result must contain.
This is the method underneath every skill in the Vault. It is not a character flaw when a technician falls into the Doom Loop... it is the default human response to pressure. The Probing Loop is the discipline that replaces it.
Five beats that feel like progress because someone is busy... opening panels, pulling parts, running around... but at no point does it generate new information about the system.
Three steps, always in this order, and it is a loop, not a checklist. Every pass generates real information and converges on the actual cause.
Fifteen qualified skills, organized into four tiers, plus the application, records, workbooks, cases, guides, and course that turn the methods into practice.
Standard-library Python, local SQLite storage, CSV export, and source-gated status.
Four blank DOCX templates with four matched worked examples.
Blank and worked battery-capacity files with user-entered criteria and verified formulas.
Real troubleshooting movement rebuilt around evidence, authority, and recovery.
IR evidence, field-case practice, GMDSS battery workbooks, and 4-20 mA loops.
Trace the Circuit in editable PPTX and delivery PDF formats.
The reasoning layer every field action builds on.
How to run a live fault through Identify, Probe, Steer instead of the Doom Loop of swapping parts under pressure.
Follow a control circuit from source to return, predict the next state, and choose the measurement that separates competing causes.
Measurement methods with explicit context and source boundaries.
Proper megger work: isolation, discharge, test voltage selection, PI and DAR ratios.
Proper multimeter method, including ghost voltage, loaded versus unloaded readings, and proving dead correctly.
Interpret insulation readings through temperature, trend, test conditions, and the criterion that actually governs the asset.
Plan and document thermal surveys without turning an image or generic temperature difference into unsupported severity.
Prepare, log, calculate, and verify a capacity test against the selected maker, SMS, flag, or Class/RO source.
Separate sensor, wiring, input, scaling, and actuator faults with source-to-return measurements and recovery proof.
Turning field data into a report that hands over cleanly.
Drafting a periodic vessel electrical audit report, with the generic template included.
Insulation test reports with a proper readings register and judgment column.
Thermal survey reports with image pairs, traceable evidence, and source-backed finding classification.
Short and emergency visit reports that hand over cleanly to the next person.
Building the system around the skills.
Run the tested offline evidence logger for IR, SPM, vibration, and battery records without universal pass values.
Set up a Notion maintenance workspace and connect its records to a compatible AI workflow.
Run several AI sessions as one organized team when preparing documentation or working a big job.
The worked examples show completeness, not acceptance values to copy. Every field criterion still comes from the applicable maker instruction, vessel procedure, permit, standard, flag, Class/RO requirement, and responsible authority.
ProbeTrack is the working application inside SkillVault. It runs locally with standard-library Python and SQLite, preserves the evidence behind a reading, and judges only against a criterion source you selected.
Stores spot readings, PI, DAR, test conditions, and user-sourced correction factors without inventing a universal megohm limit.
Retains dBm, dBc, the arithmetic difference, and the device indication without pretending to reproduce proprietary normalization.
Preserves velocity, measurement point, mounting, load, and operating context without forcing one machine class onto every asset.
Calculates delivered ampere-hours and capacity percentage, then compares only with the selected maker, SMS, flag, or Class/RO criterion.
The status says what happened against that named source. It does not turn the source into a universal rule.
The evidence stays useful while ProbeTrack tells the truth about what remains unresolved.
Still building the reflexes for live fault work. The Probing Loop gives you the discipline before the experience does, and the reporting skills teach you what a finished document is supposed to look like.
You know your way around a panel. What you need next is the method for the faults that do not have an obvious answer, and the reporting habits that make you look ready for more responsibility.
Strong on mechanical systems, less confident on the electrical side. These skills walk you through the testing method and the interpretation, not just the steps.
What is included, how the tools work, and where software stops and professional authority begins.
You receive the complete ProbeLem SkillVault v1 as one digital ZIP: 15 reusable AI skills across four tiers; the tested ProbeTrack 1.1 Python and SQLite application; four editable DOCX report templates with four matched worked examples; two GMDSS battery-capacity XLSX workbooks; eight anonymized maritime electrical field cases; four guide PDFs; and the editable 30-slide Trace the Circuit course with its PDF copy.
The package also includes Start Here, safety and scope, contents, license, version notes, application tests, and a SHA-256 integrity manifest. The worked examples demonstrate complete records, not universal acceptance values to copy into field work.
Read START_HERE.md and
SAFETY_AND_SCOPE.md. Then open the Trace the Circuit PDF
and work through Case 1 in the Field Case Evidence Workbook. For a
live supervised task, begin with the Probing Loop skill. For a
schematic-driven fault, begin with Trace Control Circuits. Use
ProbeTrack when you need structured local readings and exports.
Choose the closest skill, give the assistant the observed symptom,
asset boundary, available drawings, authorized test state, and the
deliverable you need. In a chat workspace, attach the selected
SKILL.md file and its references. In a file-aware coding
assistant, keep the complete skill folder intact so its templates and
references remain connected.
No. Each skill is a method framework, not a phrase you paste in. The skills carry trigger conditions, safety gates, an evidence method, worked examples, common mistakes, and adaptation guidance. They are designed to improve how a task is worked and checked, not merely produce one answer.
No specific AI subscription is included or required. The skills are plain markdown and work best with an AI tool that can read uploaded documents or reusable local skill folders. The quality and file-handling features available will depend on the AI service you choose.
ProbeTrack 1.1 is the local evidence logger included with the package. It stores assets, test context, raw and derived readings, criterion sources, and status results in a SQLite database. It supports insulation resistance, SPM, vibration, and battery-capacity records, with review and CSV export commands.
ProbeTrack requires Python 3.10 or later and terminal access. It uses only Python's standard library, needs no third-party packages, and can operate offline. The remaining package uses common formats: Markdown, DOCX, XLSX, PPTX, PDF, CSV, and SQLite.
No. ProbeTrack is the tested local Python and SQLite application. The ETO Notion Workspace is a separate skill for designing related maintenance, evidence, job, spare, and handover records in Notion. You can use either workflow without treating the software as the acceptance authority.
A reading does not become acceptable or unacceptable by itself.
ProbeTrack assigns RECORDED_NOT_JUDGED when no applicable
criterion profile and source have been selected. The evidence remains
stored, but the program refuses to manufacture a universal limit.
No. ProbeTrack records evidence and can compare a result only with
an entered criterion tied to a named source. If that source or limit
is absent, it returns RECORDED_NOT_JUDGED. Release to
service remains with the responsible human authority.
The package teaches transferable evidence and troubleshooting methods, not one make or model. Adapt each workflow using the exact equipment manual, drawings, vessel procedures, operating state, and governing maker, SMS, flag, or Class/RO criteria. Worked-example values demonstrate completeness and must not be copied as field acceptance limits.
No. This is educational and workflow material for qualified or supervised personnel. It does not replace competency requirements, permits, isolation, lockout/tagout, prove-dead procedures, personal protection, maker instructions, or the responsible officer's authority.
The purchaser may use and adapt the files for personal learning, work, and internal documentation. Resale, redistribution, sublicensing, publishing the files as a competing library, commercial team use, training-provider use, and white-label use require separate written permission from ProbeLem.
No. This is an independent educational product built from public engineering principles, original training material, anonymized field lessons, and Lem's experience as a traveling Fleet ETO. It is not endorsed by any employer, vessel operator, shipping company, maker, flag, or Class/RO organization.
The release passed 397 automated package checks with zero errors. All 15 skills passed their structural scorecard, all five ProbeTrack tests passed, and the DOCX, XLSX, PPTX, PDF, SQLite, CSV, archive, and integrity-manifest paths were checked as one delivery system.
Contact connect@probelem.com or WhatsApp Business. Lem works at sea, so support is answered after working hours... expect a reply within 24 to 48 hours rather than instantly.
Fifteen skills, ProbeTrack, field tools, worked examples, and one evidence-first method underneath them. Put the Probing Loop to work on your next supervised troubleshooting or documentation task.