Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Initialize sources into the processing pipeline. Takes URLs, files, or raw text and creates inbox items with full provenance metadata. The entry point for all external knowledge — nothing enters the system without being seeded first. Triggers on: "seed", "add source", "ingest url", "new input"
.claude/skills/miosa-osa-seed/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-16 | ✗→✓ | ▲ Improved | 25% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 3% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 342% | 0% |
| case-04 | ✗→✓ | ▲ Improved | 51% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -44% | 0% |
> Initialize sources into the processing pipeline with provenance metadata.
The entry point for all external knowledge. Takes raw inputs — URLs, files, pasted text, clipboard — and creates structured inbox items with provenance tracking (where it came from, when, who authored it, why it was added). Nothing enters the knowledge system without being seeded first. This ensures every piece of knowledge has a traceable origin.
bash# Seed a local file /seed path/to/article.md # Seed a URL (fetches and converts) /seed --url "https://example.com/blog-post" # Seed inline text /seed --text "Ed mentioned that enterprise pricing should be $2K/seat" # Seed with explicit metadata /seed path/to/notes.md --source "Team sync call" --author "Roberto" --reason "Pricing decision" # Seed multiple files /seed path/to/inbox/*.md --batch # Seed from clipboard /seed --clipboard
| Flag | Type | Default | Description | |------|------|---------|-------------| | <input> | positional | — | File path or glob pattern | | --url | string | — | URL to fetch, convert to markdown, and seed | | --text | string | — | Inline text to seed | | --clipboard | flag | false | Read from system clipboard | | --source | string | auto-detect | Source name (publication, meeting, person) | | --author | string | unknown | Author of the source material | | --date | date | today | Date of the source material | | --reason | string | — | Why this is being added (aids future retrieval) | | --tags | string] | auto | Initial topic tags | | --priority | enum | normal | high, normal, low — affects processing order | | --batch | flag | false | Process multiple inputs as separate seed items | | --output-dir | path | inbox/ | Where to write seeded items |
/reduce).yaml--- id: seed-2026-03-20-143000-a1b2c3 type: seed status: inbox source: "https://example.com/ai-pricing-strategies" source_title: "Enterprise AI Pricing in 2026" source_author: "Jane Smith" source_date: 2026-03-18 acquired_at: 2026-03-20T14:30:00Z acquired_via: url_fetch content_hash: sha256:a1b2c3d4... word_count: 2847 language: en tags: [pricing, enterprise, AI, SaaS] reason: "Research for AI Masters pricing decision" priority: normal --- # Enterprise AI Pricing in 2026 [Full converted content here...]
--url mode)--clipboard mode)| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-14 | pass→pass | 8,187 | 2,510 | -69% | 1 | 1 | 0% | 1,249 | 1,266 | +1% | 0 | 0 | — |
case-15 | pass→pass | 20,706 | 5,823 | -72% | 1 | 1 | 0% | 3,110 | 2,057 | -34% | 0 | 0 | — |
case-16 | fail→pass | 17,371 | 16,849 | -3% | 1 | 1 | 0% | 2,985 | 3,734 | +25% | 0 | 0 | — |
case-01 | fail→fail | 6,304 | 6,265 | -1% | 1 | 1 | 0% | 1,127 | 2,143 | +90% | 0 | 0 | — |
case-02 | fail→pass | 12,225 | 8,553 | -30% | 1 | 1 | 0% | 2,183 | 2,256 | +3% | 0 | 0 | — |
case-03 | fail→pass | 3,309 | 6,306 | +91% | 1 | 1 | 0% | 433 | 1,914 | +342% | 0 | 0 | — |
case-04 | fail→pass | 7,764 | 4,437 | -43% | 1 | 1 | 0% | 1,058 | 1,596 | +51% | 0 | 0 | — |
case-05 | fail→pass | 17,690 | 3,210 | -82% | 1 | 1 | 0% | 2,481 | 1,381 | -44% | 0 | 0 | — |
case-06 | fail→pass | 14,685 | 9,004 | -39% | 1 | 1 | 0% | 2,596 | 2,717 | +5% | 0 | 0 | — |
case-07 | fail→pass | 20,020 | 12,382 | -38% | 1 | 1 | 0% | 3,079 | 2,947 | -4% | 0 | 0 | — |
case-08 | fail→pass | 10,324 | 2,153 | -79% | 1 | 1 | 0% | 1,345 | 1,324 | -2% | 0 | 0 | — |
case-09 | fail→pass | 5,334 | 5,561 | +4% | 1 | 1 | 0% | 800 | 2,040 | +155% | 0 | 0 | — |
case-10 | fail→pass | 12,632 | 3,685 | -71% | 1 | 1 | 0% | 1,566 | 1,478 | -6% | 0 | 0 | — |
case-11 | pass→pass | 17,301 | 22,545 | +30% | 1 | 1 | 0% | 2,416 | 4,174 | +73% | 0 | 0 | — |
case-12 | pass→pass | 17,333 | 8,506 | -51% | 1 | 1 | 0% | 2,902 | 2,588 | -11% | 0 | 0 | — |
case-13 | fail→pass | 5,864 | 1,704 | -71% | 1 | 1 | 0% | 913 | 1,203 | +32% | 0 | 0 | — |
case-17 | fail→pass | 11,549 | 2,304 | -80% | 1 | 1 | 0% | 1,745 | 1,302 | -25% | 0 | 0 | — |
case-18 | fail→fail | 7,391 | 5,701 | -23% | 1 | 1 | 0% | 1,169 | 1,747 | +49% | 0 | 0 | — |
case-19 | fail→pass | 11,009 | 9,982 | -9% | 1 | 1 | 0% | 1,716 | 2,576 | +50% | 0 | 0 | — |
case-20 | pass→pass | 16,361 | 10,590 | -35% | 1 | 1 | 0% | 2,798 | 2,619 | -6% | 0 | 0 | — |
case-21 | fail→pass | 3,456 | 12,252 | +255% | 1 | 1 | 0% | 517 | 2,636 | +410% | 0 | 0 | — |
case-22 | fail→pass | 11,762 | 4,613 | -61% | 1 | 1 | 0% | 464 | 1,900 | +309% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted, and 21 counted toward the lift figure. The other 1 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +68 percentage points is the difference between those two pass rates over the 21 comparable cases.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.