verify-sources

name	verify-sources
description	Captures human source verification for tracks, timestamps it, and updates track files. Use when sources need human review before generation.
argument-hint	<album-name>
model	claude-sonnet-4-6
allowed-tools	["Read","Edit","Glob","Grep","Bash"]

Your Task

Input: $ARGUMENTS (album name)

Guide the user through source verification for all tracks with pending sources in the specified album.

---

Source Verification Skill

You facilitate the human source verification gate — the critical checkpoint between research and generation. No track should be generated on Suno until a human has verified that all sources are real, accessible, and accurately represented.

---

Step 1: Find the Album

1. Call find_album(name) — fuzzy match by name, slug, or partial
2. If not found, MCP returns available albums

Step 2: Identify Pending Tracks

1. Call get_pending_verifications() — returns all pending tracks grouped by album
2. Filter to the target album

If no pending tracks:

All tracks in [album] have been verified. No action needed.

If pending tracks exist, list them:

SOURCE VERIFICATION: [Album Title]
===================================

Tracks needing verification:
  1. [track-slug] — [track-title]
  2. [track-slug] — [track-title]
  ...

Total: X tracks pending verification

Step 3: Walk Through Each Track

For each pending track:

1. Call extract_links(album_slug, track_slug) — extracts markdown links from the track file
2. Call extract_links(album_slug, "SOURCES.md") — get the full citation list
3. Read RESEARCH.md (if it exists) for evidence chains, confidence levels, and claim-to-source mappings — this gives the human verifier context for what each source is supposed to support, not just the URL
4. Present sources to the user:

TRACK: [track-title]
--------------------
Sources referenced in this track:

  1. [Source Name](URL) — [brief description of what it supports]
  2. [Source Name](URL) — [brief description]
  ...

Please verify:
  - Each URL is accessible and contains the claimed information
  - No sources are fabricated or hallucinated
  - Claims in lyrics are supported by cited sources

Type "verified" to confirm, or describe any issues.

4. Wait for user response:
   • If "verified" (or equivalent affirmative) → update the track
   • If issues reported → note them, ask user how to proceed

Step 4: Update Track Files

When user confirms verification for a track:

1. Call update_track_field(album_slug, track_slug, "sources-verified", "✅ Verified (YYYY-MM-DD)") — updates the field and auto-rebuilds state cache

   • Use today's date

2. Confirm the update:

✅ [track-title] — Sources verified (2025-02-06)

3. Move to next pending track

Step 5: Update Album Status

After all tracks are verified:

1. Check if album status should advance:

   • If album was Research Complete → update to Sources Verified
   • If album was In Progress and all tracks now verified → note it

2. Rebuild state cache: Call rebuild_state() to ensure MCP server has fresh data

3. Summary report:

VERIFICATION COMPLETE
=====================
Album: [title]
Tracks verified: X/Y
Date: YYYY-MM-DD

All sources verified. This album is cleared for lyric writing.
Next step: /bitwize-music:lyric-writer [track] (write lyrics from verified sources)

---

Handling Issues

If the user reports a problem with a source:

1. Document the issue in the track file as a comment or note
2. Do NOT mark as verified — keep status as Pending
3. Suggest resolution:
   • Source URL broken → "Can you find an updated URL?"
   • Source doesn't support claim → "Should we revise the lyric, or find a supporting source?"
   • Source is fabricated → "I'll remove this source. Do we need to revise the track?"
4. After resolution, re-present for verification

---

Remember

• Never auto-verify — this skill exists specifically for human review
• Present sources clearly — the user needs to actually check each URL
• Date-stamp everything — verification dates matter for audit trail
• One track at a time — don't rush through, each track matters
• Update state cache — after changes, run indexer update so MCP server has fresh data