← All agent skills

Campaign reporting skill for your AI assistant

Give your assistant the post URLs. Get public metrics, availability checks and a dated snapshot to compare with your next reading.

Free to install. Sourcevine API usage consumes credits.

For Codex and Claude Code Requires Sourcevine MCP Download the skill file ↓

TikTokInstagramYouTubeX

An example report.

Download example Markdown

Recommended next move · Fictional example

Resolve the missing Instagram reading before sending a complete delivery update.

Three of four posts have current metrics. Two comparable view counts increased; one decreased.

No campaign targets, conversion data or paid-delivery evidence were supplied.

Open the complete example report ↗
Check firstEvidence: fictional-tiktok-1, fictional-instagram-2, fictional-youtube-3, fictional-x-4

Confirm the reel URL with the campaign owner

The lookup returned not found. That does not establish deletion or failed delivery.

Proposed copy · Verify before use

Three of four posts have current metrics. TikTok views rose by 2,000 and YouTube by 500; X fell by 200 between the recorded readings. The Instagram lookup returned not found and needs a link check. These observations do not establish conversions or paid delivery.
See the action plan
  1. Confirm that the supplied link is the intended deliverable.
  2. Record any corrected URL for an explicitly requested next run.
What to learn

Resolve the missing reading or document the remaining gap in the next update.

Confirm first
  • Intended reel URL
  • Whether any contractual target applies
MonitorEvidence: fictional-x-4

Keep the X decrease visible

Views changed from 2,600 to 2,400 between comparable readings; the cause is unknown.

Proposed copy · Verify before use

X shows a decrease of 200 views between readings. We are preserving that change without attributing a cause.
See the action plan
  1. Keep both timestamps and values in the client update.
  2. Use the saved snapshot for a later authorized comparison.
What to learn

Check whether a later comparable reading changes the picture; do not infer fraud or campaign impact.

Confirm first
  • Preferred time for the next reading

October launch

4 posts · Comparison with the previous snapshot

Fictional example

Readings: 3 → 4 October 2026, 09:00 UTC. No real posts were fetched for this example.

3 available1 not foundUp to 5 credits at live prices
Fictional public post metrics, grouped by platform. View changes compare two collection times.
Post / platformStatusViewsView changeLikesComments / X replies
Launch videoTikTokAvailable12,000+2,00042032
Product reelInstagramNot foundNot returnedNo comparable readingNot returnedNot returned
Product walkthroughYouTubeAvailable8,500+500280Not returned
Launch postXAvailable2,400-200648
  • Launch video

    TikTok

    Available
    Views
    12,000
    View change
    +2,000
    Likes
    420
    Comments
    32
  • Product reel

    Instagram

    Not found
    Views
    Not returned
    View change
    No comparable reading
    Likes
    Not returned
    Comments
    Not returned
  • Product walkthrough

    YouTube

    Available
    Views
    8,500
    View change
    +500
    Likes
    280
    Comments
    Not returned
  • Launch post

    X

    Available
    Views
    2,400
    View change
    -200
    Likes
    64
    Replies
    8

Gaps stay visible. Not-found posts have no invented counts. Negative changes stay visible. Views are not unique reach or proof of paid delivery.

Credits used. This example would use up to 5 credits at current live prices. Real reports include the charges returned by the tools.

Download complete HTML report ↓

Try a comparison with these fictional snapshots. No API calls needed.

Previous snapshot ↓Current snapshot ↓How to try the example →

How the skill works.

  1. 01

    Connect your assistant

    Set up Sourcevine MCP and install the skill in Codex or Claude Code.

  2. 02

    Add your campaign links

    Provide public post URLs and a credit limit. The skill estimates the cost before reading.

  3. 03

    Keep the snapshot

    Read the report now. Supply the saved snapshot next time to compare observed changes.

Add Campaign Report to your assistant.

Choose your client and run its command on macOS or Linux. For live data, connect Sourcevine MCP once. The fictional examples work offline.

  1. 1. Confirm your account and save a key. Start setup →
  2. 2. Connect Sourcevine MCP. Open connection setup →
  3. 3. Install this skill. Choose your assistant below and copy its command.
  4. 4. Try the starter prompt. Copy your first request →
Codex
Personal skill · v1.1.0
skill_dir="$HOME/.agents/skills/sourcevine-campaign-report"
mkdir -p "$skill_dir" &&
curl -fL "https://sourcevine.io/skills/sourcevine-campaign-report/SKILL.md" \
  -o "$skill_dir/SKILL.md.download" &&
mv "$skill_dir/SKILL.md.download" "$skill_dir/SKILL.md"

After installation, start a new session if needed and invoke $sourcevine-campaign-report.

Claude Code
Personal skill · v1.1.0
skill_dir="$HOME/.claude/skills/sourcevine-campaign-report"
mkdir -p "$skill_dir" &&
curl -fL "https://sourcevine.io/skills/sourcevine-campaign-report/SKILL.md" \
  -o "$skill_dir/SKILL.md.download" &&
mv "$skill_dir/SKILL.md.download" "$skill_dir/SKILL.md"

After installation, start a new session if needed and invoke /sourcevine-campaign-report.

Prefer a manual download? Save SKILL.md in a folder named sourcevine-campaign-report in your client’s skill directory. Windows, updates and troubleshooting →

A first request to try.

After setup, paste this into your assistant and add your campaign links.

Use the Sourcevine campaign report skill for these public post URLs:
[paste your campaign post URLs here]

Use at most 10 credits.
Lead with the decision, key changes and prioritized follow-ups.
Save the HTML report, Markdown report and JSON snapshot.
If any posts cannot be checked, include them with an explanation.

Each run saves a snapshot for the next one.

Start with a snapshot. Add a comparison when you have another reading.

  • Checks up to 10 unique posts by default, with a 20-credit estimated limit.
  • Preserves missing metrics, private results and collection timestamps.
  • Uses cached readings by default; request a fresh read when you need one.
  • Runs when you ask. Recurring schedules are a separate setup.