Use before expensive VidSeeds MCP workflows - seeds, daily MCP call quotas, async jobId polling, when to call vidseeds_get_seed_balance, and avoiding retry loops on MCP_QUOTA_EXCEEDED or insufficient seeds.
SkillScore73Maintained
Automatic
◆Marketing & Content pack
Last verified 8/15/2026
What it changes
Passed 12 of 16 test cases with the skill, and 4 of 16 without it.
Without skill27%
With skill73%
Checks it passed that the model alone missed
Identifies missing Personal Access Token as the cause.
Polls the job status continuously until completed or failed.
Directs checking seed balance using vidseeds_get_seed_balance.
and 9 more in the full report
10 fixed6 passed either way6 still failing0 broke
Pass-rate lift
+45 pp
Tokens
+29%
Agent turns
0%
Response time
-61%
Model gemini-3.6-flashMeasured 8/15/2026v116 of 22 cases passed with skill · 16 comparable
Tokens, turns and time describe resource use, not a dollar price.
See it in action
One of 10 test cases the skill fixed. Same model, same prompt, with and without the skill.
1 of 10
Prompt
Does the Sprout plan daily quota of 1,000 MCP calls reset completely at midnight UTC every night?
What was checked, without and with the skill
Checked
Without skill
With skill
Clarifies that call quota uses continuous refill rather than a midnight reset.
Missed
Passed
case-18 · gemini-3.6-flash · 8/15/2026 · answers are capped and redacted
What's inside
The instructions your model reads when the skill loads.
SKILL.md69 linessingle file
VidSeeds MCP efficiency
Read this before multi-step or charge-bearing workflows.
Orient first (free)
New to this server, or unsure which tool does the job? Call vidseeds_guide - it returns the capability map, the recommended tool chain for a stated goal, and what VidSeeds does not do. It is free and quota-exempt, so it never costs seeds or a call-bucket token. The server's initialize instructions and the vidseeds://guide resource carry the same orientation; workflow prompts (e.g. optimize-youtube-video, make-thumbnail) hand you the ready-made tool chain. Reach for these before concluding a capability is missing. Parameter-level field docs are intentionally omitted from tools/list JSON schemas to save context - use vidseeds_guide (with a goal) or read vidseeds://guide when you need argument semantics beyond types/enums.
Two cost layers
Per-tool seeds - Some tools spend seeds (thumbnails, intelligence, translation, etc.). Read-only tools are free. Charge-bearing tools state their seed cost in the first sentence of their compressed tools/list description; when it is not there, call vidseeds_guide with the tool name for the confirmed cost. Confirm with the user before charge-bearing calls. (2026-06: use fields: ["description","tags"] on regenerate for light/correct-price partials; thumbnail edits cost more than generation when using input reference image.)
Daily MCP call quota - Almost every tool call counts toward a plan bucket (continuous refill, not midnight reset). When the bucket is empty, each extra call costs 1 seed on top of any per-tool cost.
Plan
Refill / day
Bucket cap
Sprout
1,000
2,000
Growth
3,000
6,000
Harvest
8,000
16,000
Agency
20,000
40,000
Trial
1,000
1,000
Free and quota-exempt:vidseeds_get_seed_balance, vidseeds_get_seed_balance_and_subscription, vidseeds_guide.
Before expensive work
vidseeds_get_seed_balance - seeds, remaining included MCP calls, overage rate (free).
Tell the user expected per-tool seed cost (from the tool's compressed description, or vidseeds_guide when the description omits it).
Prefer async + poll for long jobs instead of hammering sync tools that may time out at the edge.
Promo campaign note: vidseeds_generate_promo_campaign_pack is seed-charged on successful AI generation; conversion to drafts with vidseeds_create_social_posts_from_campaign_pack is not the billable AI step.
Async job pattern
Many writes return immediately with jobId (and often pollTool or a named poll tool in the response):
Call the async or generate tool → capture jobId and expectedCostSeeds if present.
Poll the indicated tool every ~3s until completed or failed.
On failed, surface the error once; do not retry in a tight loop.