Novia-RDI-Seafaring

File a structured Agent Experience (AX) friction report when a tool you are using trips you up. Use this the moment a tool's contract does not match reality — its docs/skill claim a capability that is missing or named differently, a command mixes logs into machine output, a check or probe reports failure when the real operation works, an error is an opaque traceback instead of a structured message, or a result forces a needless second call. Report observed friction with evidence (what you expected, what happened, what you did instead), not wishes. Also files lower-priority feature requests, kept clearly separate. Reports are written locally for a human to triage. Default to filing the moment friction happens; for a deliberate evaluation of a tool, see "Running an evaluation session" below. Trigger words: "AX", "agent experience", "this tool tripped me up", "the docs were wrong", "report friction", "file an AX report", "AX-test this tool", "evaluate the agent experience".

Run a valid, hard-to-cheat Agent Experience (AX) test by driving a REAL harness as an unbiased test subject. Use when you (the orchestrator) want to evaluate a tool's AX with a fresh agent, without that agent cheating by reading the tool's source, and without your own context biasing it. Sets up an isolated workspace, launches a real harness, keeps the run valid, then hands the reports to ax-review. Trigger words: "run an AX test", "validate the AX", "AX-test a tool with a clean harness", "set up an AX evaluation", "test the agent experience without cheating".

Triage Agent Experience (AX) reports: read the filed reports, group them, summarize the friction, and decide what to promote into a tool's issue tracker. Use when asked to "triage AX reports", "review the ax-reports", "what AX issues were filed for <tool>", "summarize agent friction", or "what should we fix first". You are the reviewer half of AX reporting — the reports are written by the ax-report skill. Trigger words: "triage", "read the AX reports", "review AX reports", "AX findings", "what friction was reported".

Use this skill when the user works with engineering documents — datasheets, leaflets, manuals, P&ID drawings — and wants to ingest them into a source-grounded knowledge base, query the structured contents, or build a workspace canvas where every value points back to its source page + bbox. ANCHOR has its OWN extraction pipeline (docling layout + OCR → per-page markdown and images → structured regions with bboxes). So when the user says "ingest this PDF", "read / OCR this document", "extract the specs", "what does the leaflet say about X", "place that spec on the canvas", or "wire this value into a simulation", drive it through ANCHOR — do NOT install OCR/PDF libraries or write your own parsing code.

Run an FMPy simulation for a given FMU file. Use when the user wants to simulate an FMU, run a Functional Mock-up Unit, or work with FMI-based models. Also handles follow-up messages from the simulation config widget (messages starting with "SIMULATE:").