Structured search and file results
search keeps its console tables and CSV output. In AI/MCP execution it can also return a data object containing machine-readable records. fileinfo uses the same result model for exact-file inspection.
Read a record
Use fullPath for subsequent commands rather than reconstructing it from a filename. File records include fileId, folderId, and available version, state, revision and checkout metadata. Folder records use itemType: "folder" and have no file ID. Requested search variables appear when available.
Null means unavailable, not false. Read each item's warnings before treating a value as authoritative.
Check completeness
| Field | Meaning |
|---|---|
schemaVersion |
Currently 1. |
kind, items |
Result type and captured records. |
complete |
Whether the operation completed; it does not promise every record fits in the snapshot. |
matchedCount, returnedCount, omittedCount |
Observed counts, not an estimated total for the vault. |
truncated |
The snapshot omitted records. |
captureFailures, incompleteReason |
Capture or execution problems that need review. |
Search snapshots retain at most 2,000 records. Narrow a search if records were omitted. Interrupted, license-limited or failed captures must not be described as a complete vault inventory.
Retrieve paged output
When MCP reports outputTruncated: true, use pdmshell_read_result and concatenate its text pages to recover the retained, redacted JSON. snapshotTruncated means the complete original result was not retained and cannot be reconstructed by paging.
Do not rerun a mutating command merely to retrieve its output. Read the retained result instead. Search and file inspection do not authorize later edits to the returned files.