Cover Letter Generator

Batch process multiple JDs by triggering the n8n cover-letter-generator workflow via MCP tools.

Workflow

1. Parse Input File

Excel (.xlsx/.xls):

bash

python scripts/parse_input.py jobs.xlsx

•Auto-detects columns: Job Title/Position, Full JD/Job Description, Company
•Filters for PASS if Filter Result column exists
•Outputs JSON array to stdout

Text (.txt):

bash

python scripts/parse_input.py jds.txt "CompanyName"

•Format: sections separated by ---, with COMPANY:, POSITION:, JD: headers
•Second arg is default company name

Output format:

json

[{"company": "X", "position": "Y", "jd": "..."}]

2. Find the Workflow

Use n8n MCP to locate the cover-letter-generator workflow:

code

n8n_list_workflows → search for "cover-letter" or "Cover Letter Generator"

Note the workflowId for triggering.

If workflow not found, inform user to check n8n.

2.5. Compute Unique Keys

Keys MUST be globally unique across all batches. Read the counter file to determine starting key:

bash

# Read next available key (default 0 if file doesn't exist)
cat exp_lib/.next_key 2>/dev/null || echo 0

Assign keys sequentially: if starting key is 15 and batch has 3 jobs, keys are 15, 16, 17.

After generating the order file (Step 4), update the counter:

bash

echo <next_unused_key> > exp_lib/.next_key

Example: batch used keys 15-17, write 18 to .next_key.

3. Trigger Workflow for Each JD

CRITICAL: Use curl in background, NOT n8n_test_workflow MCP tool.

The webhook uses responseMode: lastNode and takes ~85 seconds to complete. The MCP n8n_test_workflow tool has a 30-second timeout and will report "No response" even though the workflow was successfully triggered. This leads to duplicate triggers and wasted API credits.

For each job, trigger via curl in the Bash tool (run_in_background=true):

bash

curl -X POST "http://localhost:5678/webhook/cover-letter-generator" \
  -H "Content-Type: application/json" \
  -d '{"jobDescription": "<full JD text>", "key": <globally_unique_key>}' \
  --max-time 300

Important:

•Use the globally unique key from Step 2.5, NOT a 0-based index
•Wait 30 seconds between each trigger to avoid overwhelming n8n
•Report progress: [1/N] Key=15 | Company - Position... Sent
•NEVER retry if curl times out — check execution status first (see Step 3.5)
•If a trigger fails, check n8n_executions before retrying

3.5. Wait for Completion & Extract PDFs

After triggering all JDs, wait ~90 seconds per job, then verify completion:

javascript

n8n_executions({
  action: "list",
  workflowId: "<workflow-id>",
  limit: <number_of_jobs_sent>
})

Confirm all executions show status: "success". If any show status: "error", check details before retrying.

Extract PDFs from Docker container (PDFs are generated inside n8n-n8n-1 container):

bash

docker cp n8n-n8n-1:/home/node/.n8n-files/data/<key>-R.pdf exp_lib/<key>-R.pdf
docker cp n8n-n8n-1:/home/node/.n8n-files/data/<key>-CL.pdf exp_lib/<key>-CL.pdf

Do this for each key. Verify file sizes are non-zero.

4. Generate Order File

After all JDs are sent, write the order file using the Write tool to exp_lib/batch_order_<timestamp>.json:

json

{
  "batch_id": "2026-01-27_16-45-00",
  "created_at": "2026-01-27T16:45:00",
  "total_jobs": 2,
  "interval_seconds": 30,
  "jobs": [
    {
      "key": 15,
      "order": 1,
      "sent_at": "2026-01-27T16:45:00",
      "company": "Company A",
      "position": "Job Title A",
      "jd": "...",
      "status": "sent",
      "expected_files": {
        "resume": "15-R.pdf",
        "cover_letter": "15-CL.pdf"
      }
    },
    {
      "key": 16,
      "order": 2,
      "sent_at": "2026-01-27T16:45:30",
      "company": "Company B",
      "position": "Job Title B",
      "jd": "...",
      "status": "sent",
      "expected_files": {
        "resume": "16-R.pdf",
        "cover_letter": "16-CL.pdf"
      }
    }
  ]
}

5. Report Summary

Print summary:

•Total sent / failed count
•Order file path
•Remind user: wait ~2-3 min per job for n8n to generate PDFs
•Next step: use job-application-tracker skill to organize files

Notes

•DO NOT use n8n_test_workflow — it will timeout and mislead you into retrying
•30-second interval between triggers prevents overwhelming n8n's async processing
•Each trigger starts async generation (~85 seconds per JD)
•Order file enables correct file-to-position matching
•PDF naming: {key}-R.pdf (resume), {key}-CL.pdf (cover letter)
•PDFs are generated inside Docker container n8n-n8n-1 at /home/node/.n8n-files/data/
•Must use docker cp to extract PDFs to local exp_lib/ before organizing
•Workflow ID for Cover Letter Generator: O77JGewQqhwqvBdZ2LzYa