| name | wiki-ingest-next-batch |
| description | Use when continuing a batch import that was started in another session — when the user says "ingest next batch" or opens a parallel session during a multi-batch import. |
Knowledge Base - Ingest Next Batch
This skill handles Sessions 2–N of a multi-batch import started by wiki-ingest.
Step 1 — Check state
Before claiming a batch, verify the import is in progress:
ls .import/batch-import-[0-9]*.txt 2>/dev/null | grep -v '\.claimed\.' | sort -V
ls .import/batch-log-*.jsonl 2>/dev/null
- No unclaimed
.txt files AND no .jsonl log files: nothing to do — tell the user "No batch import in progress. Use wiki-ingest to start a new import."
- No unclaimed
.txt files BUT .jsonl log files exist: all batches are processed. Tell the user: "All batches appear complete. Say finalize ingest (or /wiki-finalize-ingest) to merge logs and rebuild indexes."
- Unclaimed
.txt files exist: proceed to Step 2.
Step 2 — Claim a batch atomically
Run the following to find and claim the next unclaimed batch:
for f in $(ls .import/batch-import-[0-9]*.txt 2>/dev/null | grep -v '\.claimed\.' | sort -V); do
mv "$f" "${f%.txt}.claimed.txt" 2>/dev/null && echo "${f%.txt}.claimed.txt" && break
done
- If the loop prints a filename (e.g.
.import/batch-import-2.claimed.txt) → that is your batch; proceed to Step 3.
- If nothing is printed → all batches are taken or already done; report "No unclaimed batch found — another session may be processing the last batch" and stop.
Step 3 — Process the batch
Determine the batch number from the claimed filename (e.g. .import/batch-import-2.claimed.txt → N=2). Dispatch sub-agents in batches of 10 when the current client supports and permits delegation; otherwise process the files in the current session in manageable chunks. Each sub-agent prompt must begin with: "Invoke wiki-ingest-per-note before processing. Write session logs to .import/batch-log-N.jsonl. Then ingest these files: [list]." (Replace N with the actual batch number.)
Step 4 — Finish
After processing all files, delete the .claimed.txt file.
Report:
- Batch number claimed
- Number of notes processed
- Pages created and updated
Then check if more unclaimed batches remain:
ls .import/batch-import-[0-9]*.txt 2>/dev/null | grep -v '\.claimed\.'
- Unclaimed batches remain: tell the user "More batches available — other sessions can run
wiki-ingest-next-batch to claim them."
- No unclaimed batches remain AND
.jsonl logs exist: tell the user "All batches are now processed. When all parallel sessions are done, say finalize ingest (or /wiki-finalize-ingest) in the coordinator session to wrap up."
