Email Verification

1. Ask the user which email verification provider they want to use.
2. Fetch or read the provider's API documentation and identify:
   • Credit check endpoint
   • Single email validation endpoint
   • Required input fields and authentication method
   • Rate limits and request constraints
   • Response format (status, sub_status)
   • Status values and which to keep vs remove
3. Ask for their API credentials and confirm access

Step 1: Check API access and credits

1. Call the provider's credit check endpoint
2. Count unique emails in the input CSV
3. If credits < unique emails, warn the user and ask whether to proceed with partial validation or stop

Step 2: Extract and deduplicate emails

1. Read the input CSV, auto-detect the email column (common names: email, Email, email_address)
2. Build a set of unique email addresses
3. Skip rows with empty/missing email
4. Report: {N} unique emails to validate, {M} rows without email skipped

Step 3: Validate emails

For each unique email, call the provider's validation endpoint:

• Respect the provider's rate limits (from Step 0)
• Print progress every 50 emails
• Store results as {email: {status, sub_status}}
• On error (timeout, JSON parse failure), mark as unknown

Step 4: Categorize and report

Group results by the provider's status values. Use the status mapping identified in Step 0 to determine which emails to keep and which to remove.

General guidance (confirm against provider docs):

Present summary table with counts per status, then list all emails to remove with their status and sub_status.

Step 5: Output cleaned CSV

1. Filter the original CSV: keep only rows where email status maps to "keep"
2. Write cleaned CSV to same directory as input, named {original_name}_verified.csv
3. Save full validation results to {original_name}_verification_results.json for reference
4. Report: {kept} emails kept, {removed} removed, {catch_all} catch-all (flagged)

Step 6: Clean from sequencer (optional)

If the user provides sequencer campaign IDs:

1. Ask which sequencer they use (Instantly, etc.) and read its API docs from skills/campaign-sending/references/
2. For each removed email, call the sequencer's lead deletion endpoint
3. Respect rate limits
4. Report: {N} leads removed from sequencer

If no campaigns are provided, skip this step and note that the user should remove bad emails manually or re-upload with the cleaned CSV.

Output

API References

Provider API docs are fetched or read during Step 0. No pre-configured provider docs are bundled — add them to references/ as needed.