An AI skill is a structured system prompt or set of custom instructions that gives AI assistants like ChatGPT, Claude, and Gemini specialized capabilities. It includes a clear role definition, specific knowledge areas, step-by-step processes, and quality criteria that help AI perform specific professional tasks.

How does Skill This work?

Simply describe your professional expertise in plain text. Our free AI skill generator analyzes your input and transforms it into a structured system prompt. The skill is then graded on clarity, specificity, and usefulness. You can share your skill via a unique URL or copy it to use as custom instructions with any AI assistant like ChatGPT, Claude, or Gemini.

Is Skill This free to use?

Yes, Skill This is completely free. No account or payment is required. You can create unlimited skills and share them with anyone.

How is my skill graded?

Skills are graded on a scale from A+ to F (0-100 points). The grade reflects how clear, specific, and actionable your skill is. Higher grades indicate skills that AI assistants can use more effectively.

How do I use my skill with an AI assistant?

Copy your generated skill and paste it into the system prompt or custom instructions of any AI assistant. For ChatGPT, use Personalization > Custom Instructions. For Claude, add it to a Project's instructions. For Cursor or other IDEs, save it as a .cursorrules or CLAUDE.md file. The AI will then adopt that skill and perform tasks according to your expertise.

AI Skill Report Card

Scoring WIF To Packet Conversion

A-83·Oct 29, 2026·Source: Web

Quick Start14 / 15

Given an AI agent's output (a PDF packet, plus the WIF/image input it started from), score it against the 8-point rubric below and emit a scorecard.

INPUT: sample_draft.wif
AI OUTPUT: sample_draft_packet.pdf

SCORECARD
1. Draft Decoding & Validation .......... 8/10
2. Threading/Tie-up/Treadling Diagrams .. 6/10
3. Drawdown Rendering ................... 9/10
4. Yarn & Sett Calculations ............. 3/10  ⚠ BREAKS HERE
5. Draft-to-Prose Instructions .......... 5/10
6. Layout & Print-Readiness ............. 7/10
7. Historical/Source Attribution ........ 10/10
8. Sellability (branding, polish, legal) . 4/10

OVERALL: 52/80 (65%) — Usable draft, unsellable packet
VERDICT: Automation holds through visualization, breaks at commercial packaging.

Recommendation▾

Add a brief note on how to handle partial/missing inputs (e.g., no source WIF available at all) as an edge case

Workflow14 / 15

Progress:

Step 1: Identify input type (native WIF vs. scanned/photographed historical draft)
Step 2: Run or inspect the AI agent's conversion pipeline output
Step 3: Score each of the 8 sub-skills independently (don't let one bad score bias others)
Step 4: Flag the exact sub-skill where automation "breaks" (first score below 5/10)
Step 5: Write qualitative notes per category (specific, not generic)
Step 6: Produce final scorecard + verdict + recommendation
Step 7: (If video) storyboard each sub-skill as its own scored segment with visual proof

The 8 Sub-Skills (Rubric)

Draft Decoding & Validation — Did it correctly parse threading, tie-up, treadling, and warp/weft color sequences from the WIF, or correctly infer them from a scanned image (OCR/CV on grid marks, handwritten numbers, faded ink)? Deduct heavily for silent misreads (e.g., swapping a 6 for an 8).
Threading/Tie-up/Treadling Diagrams — Are the three core diagrams regenerated cleanly as vector/print-quality graphics (not blurry screenshots), correctly labeled, and consistent with the decoded data?
Drawdown Rendering — Is the woven structure simulation visually accurate (correct interlacement, color-and-weave effect), and does it match what the threading/tie-up/treadling imply?
Yarn & Sett Calculations — Does it compute warp length, yardage, sett (EPI/PPI), draw-in, and shrinkage estimates based on stated or inferred yarn size/structure? This is the most common break point — flag if the agent fabricates numbers without stating assumptions (yarn size, loom waste %, shrinkage %).
Draft-to-Prose Instructions — Does it convert the diagrammatic draft into step-by-step written weaving instructions a human could follow at the loom (warping order, tie-up sequence, treadling repeats)?
Layout & Print-Readiness — Is the PDF properly paginated, with consistent typography, margins suitable for home printing, and diagrams sized/placed sensibly (not split across pages)?
Historical/Source Attribution — If from a historical draft, does it credit the source (archive, weaver, publication date) and note any uncertainty in its transcription (e.g., "faded section reconstructed")?
Sellability — Does the final packet feel like a product: cover page, consistent branding, copyright/licensing note, difficulty rating, materials list, and no leftover placeholder text or broken image links? This is a common secondary break point.

Recommendation▾

Include guidance on scoring consistency across multiple evaluations (e.g., calibration examples for what a '5' vs '6' looks like)

Examples17 / 20

Example 1: Input: Clean, modern WIF file exported from WeavePoint for a simple 4-shaft twill. Output:

1. Decoding: 10/10 — native WIF, no ambiguity
2. Diagrams: 10/10 — clean vector regeneration
3. Drawdown: 10/10 — matches diagrams exactly
4. Calculations: 6/10 — EPI/PPI correct but no shrinkage % shown, assumes 10/2 cotton without stating it
5. Prose instructions: 8/10 — clear but misses "advance warp" reminders
6. Layout: 9/10 — clean, one diagram awkwardly split across a page break
7. Attribution: N/A (not historical) — scored 10/10 by default
8. Sellability: 7/10 — has cover page, missing licensing note
OVERALL: 70/80 (87.5%) — Strong automation performance on clean digital input.

Example 2: Input: Low-resolution scanned photo of an 1890s handwoven coverlet draft with faded handwritten tie-up numbers. Output:

1. Decoding: 3/10 — misread 2 tie-up cells, no confidence flagging on faded region
2. Diagrams: 4/10 — diagrams generated but propagate decoding errors
3. Drawdown: 4/10 — visually plausible but wrong due to upstream error
4. Calculations: 2/10 — invented yarn size (no source stated), no shrinkage estimate
5. Prose instructions: 3/10 — instructions reflect the decoding error, would produce wrong fabric
6. Layout: 6/10 — adequate but generic template, no visual connection to historical source
7. Attribution: 2/10 — no archive name, no note of damaged/reconstructed sections
8. Sellability: 3/10 — placeholder "Lorem ipsum" left in materials list
OVERALL: 27/80 (34%) — Automation breaks immediately at decoding; everything downstream inherits the error.
VERDICT: Not sellable. Requires human verification pass on handwritten historical source before any downstream step is trustworthy.

Recommendation▾

Consider a short template for the qualitative notes section so the written feedback per category is more structured, not just prose

Best Practices

Score decoding first and weight it conceptually — errors there cascade into every other category; always note in the verdict if a low decoding score explains low scores elsewhere ("inherited error" vs. "independent failure").
Demand stated assumptions for calculations (category 4). A confident wrong number is worse than a flagged uncertain one — score transparency, not just correctness.
Treat "N/A" categories fairly — non-historical sources auto-score 10/10 on attribution rather than being penalized.
Use the first sub-5 score as the declared break point in the verdict — this is the single most useful takeaway for weavers deciding whether to trust automation.
Keep scores evidence-based, citing specific diagram/page/cell references, not vague impressions.
For video format, give each of the 8 categories its own timestamped segment with a visual side-by-side (input source vs. AI output) before revealing the score.

Common Pitfalls

Don't average scores into a single number without the per-category breakdown — the diagnostic value is in where it breaks, not just the overall grade.
Don't let strong visual polish (category 6/8) inflate scores in correctness categories (1-4) — a beautiful PDF with wrong threading is still a failure.
Don't penalize historical sources for having lower raw accuracy if the agent correctly flags its own uncertainty — score honesty about limitations separately from raw accuracy.
Don't skip stating yarn/sett assumptions as fabricated-but-unstated — this is the single most common silent failure mode in category 4.
Don't treat a missing category 7 (attribution) as minor for historical sources — selling a packet without source credit is a legal/ethical issue, not just a polish issue.