By now you have a stack of validator output and a half-passed cup of coffee. The remaining skill is routing, the unglamorous habit that turns a wall of red into a list of small, scoped tasks.
Route by prefix first. `ESEF.2.x` and `ESEF.3.x` go to ESMA. `EFM.6.05.x` and `EFM.6.08.x` go to the SEC. `FR-NL-*` and `FG-NL-*` go to Dutch SBR. `UKFRC.*` goes to the UK FRC. `xbrldie:*` is XDT instance-level (`PrimaryItemDimensionallyInvalidError`, `DefaultValueUsedInInstanceError`, and friends). `xbrldte:*` is XDT taxonomy/DTS-level (`HasHypercubeMissingContextElementAttributeError`, `TooManyDefaultMembersError`). `xbrl.5.2.5.2` is calc inconsistency. Routing by prefix takes about three seconds and is the single highest-leverage move in the whole triage.
Then read the rest of the line. The message text says what the rule expects. The file path and source line tell you exactly which element to open. An auditor who opens the named file at the named line is several minutes ahead of an auditor who tries to fix code from memory.
When two codes fire together, fix the structurally earlier one first. A `xbrldte:*` (DTS-level) defect is upstream of any `xbrldie:*` (instance-level) defect that depends on it. A schema-level break upstream of a calc check tends to make the calc noise vanish once the schema heals. Fixing the wrong one first wastes a round trip through validation, and validators are not fast.
Prompts to AI agents work the same way. The prompt that says "fix all errors" produces regressions. The prompt that pastes the code, message, file, and line, names the exact element by QName and `contextRef`, and instructs the agent to leave every other attribute untouched produces a surgical patch. Surgical prompts protect the rest of the filing; broad prompts treat the whole report as a draft to be re-tagged.
The keypoints crystallise this routing reflex.
When you see a code, route by prefix: ESEF.2.x and ESEF.3.x are ESMA; EFM.6.5.x and 6.6.x are SEC; FR-NL-* and FG-NL-* are SBR; UKFRC.* is UK FRC; xbrldie:* is XDT instance level; xbrldte:* is XDT DTS level; xbrl.5.2.5.2 is calc inconsistency.
Always read the message text, file path, and source line. The most useful debugging is to open the named file at the named line.
For prompts to AI agents, paste the code, message, file, and line into the prompt context. Agents that have to guess from the code alone are less accurate.
Some codes (ESEF.2.4.1.*) carry a sub-suffix that further specialises the rule. The full code matters; do not truncate.
When two codes fire, fix the structurally earlier one first. A schema error makes downstream codes noisy and often vanishes when the structural fix lands.
3 questions. Druk op 1 tot 4 om te kiezen, Enter om te bevestigen.
Door deze quiz af te ronden, markeer je deze sectie als voltooid.
Tip: kies met 1 tot en met 4 en bevestig met Enter.