Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Specialized rules for repairing Markdown lint errors in Gold Context files (MD025, MD060, MD024).
.claude/skills/bilal140202-context-formatting-linting/SKILL.md| Model | Eval pass | Runs |
|---|---|---|
| gemini-3.6-flash | 100% | 4 |
| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-09 | ✗→✓ | ▲ Improved | 43% | 0% |
| case-12 | ✗→✓ | ▲ Improved | 104% | 0% |
| case-13 | ✗→✓ | ▲ Improved | -45% | 0% |
| case-05 | ✓→✗ | ▼ Worse | -58% | 0% |
| case-06 | ✓→✗ | ▼ Worse | -8% | 0% |
Version: Gold v1.0
Type: Utility Skill
Tier: 3 — Loaded when requested or encountering linting errors in context files
File: skills/context-formatting/SKILL.md
Primary Mode: Builder (formatting/fixing)
Purpose: Automates the correction of common Markdown linting errors generated in Gold Context Files, specifically targeting the metadata block (MD025), tight table separators (MD060), and recurring "Guidance" headings (MD024).
contexts/*.md files.When instructed to format a new context file, immediately apply these three transformations using the multi_replace_file_content tool.
The context file templates frequently arrive with metadata formatted as H1 tags at the top of the file, directly underneath the primary title. This triggers MD025 because markdown files should only have a single # H1 heading.
Action: Convert all secondary metadata hashes into bold markdown descriptors. Fix any broken line continuations.
From:
markdown**Version:** Gold v2.0 **Type:** Context File
To:
markdown**Version:** Gold v2.0 **Type:** Context File
| The context templates use compressed table alignment pipes without proper whitespace (e.g., | -------- | ------- |). The markdown linter requires spaces around the pipes for aligned column styles. |
| Action: Replace tight table separators with standard spaced separators (| --- |). Do this for EVERY table in the file. |
From:
markdown| Metric | Current Value | Target | Why It Matters | | --- | --- | --- | --- |
To:
markdown| Metric | Current Value | Target | Why It Matters | | --- | --- | --- | --- |
Context templates strategically repeat section headings like ### Guidance for Anti-Gravity or ### [Flow Name] across multiple sections. Markdown linters flag this as an MD024 collision.
Action: Suffix the duplicate heading with a parenthetical descriptor tied to its parent section, making it unique across the document.
From:
markdown## FRONTEND STACK ... ### Guidance for Anti-Gravity
To:
markdown## FRONTEND STACK (SKILL: CONTEXT FORMATTING) ... ### Guidance for Anti-Gravity (Frontend)
| 1. Do not alter the primary content. Only touch the formatting tags (#, |---, and heading names). |
multi_replace_file_content tool call containing multiple ReplacementChunks whenever possible to save time.| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-15 | fail→fail | 2,788 | 3,370 | +21% | 1 | 1 | 0% | 321 | 1,056 | +229% | 0 | 0 | — |
case-01 | fail→fail | 3,678 | 2,561 | -30% | 1 | 1 | 0% | 791 | 1,325 | +68% | 0 | 0 | — |
case-02 | fail→fail | 10,882 | 4,588 | -58% | 1 | 1 | 0% | 2,300 | 1,299 | -44% | 0 | 0 | — |
case-03 | fail→fail | 6,790 | 4,320 | -36% | 1 | 1 | 0% | 1,219 | 1,118 | -8% | 0 | 0 | — |
case-04 | pass→pass | 2,143 | 2,563 | +20% | 1 | 1 | 0% | 442 | 1,370 | +210% | 0 | 0 | — |
case-05 | pass→fail | 13,999 | 4,628 | -67% | 1 | 1 | 0% | 2,941 | 1,232 | -58% | 0 | 0 | — |
case-21 | pass→pass | 6,735 | 8,638 | +28% | 1 | 1 | 0% | 1,180 | 1,722 | +46% | 0 | 0 | — |
case-06 | pass→fail | 15,865 | 11,048 | -30% | 1 | 1 | 0% | 3,101 | 2,852 | -8% | 0 | 0 | — |
case-07 | pass→pass | 5,071 | 6,216 | +23% | 1 | 1 | 0% | 967 | 2,029 | +110% | 0 | 0 | — |
case-08 | pass→fail | 26,758 | 6,171 | -77% | 1 | 1 | 0% | 4,727 | 1,032 | -78% | 0 | 0 | — |
case-09 | fail→pass | 7,354 | 6,070 | -17% | 1 | 1 | 0% | 1,331 | 1,903 | +43% | 0 | 0 | — |
case-10 | fail→fail | 2,740 | 4,949 | +81% | 1 | 1 | 0% | 162 | 1,137 | +602% | 0 | 0 | — |
case-11 | pass→fail | 5,947 | 4,491 | -24% | 1 | 1 | 0% | 1,161 | 1,105 | -5% | 0 | 0 | — |
case-12 | fail→pass | 5,076 | 4,745 | -7% | 1 | 1 | 0% | 841 | 1,715 | +104% | 0 | 0 | — |
case-13 | fail→pass | 20,351 | 6,608 | -68% | 1 | 1 | 0% | 3,736 | 2,042 | -45% | 0 | 0 | — |
case-14 | fail→fail | 15,024 | 8,808 | -41% | 1 | 1 | 0% | 2,841 | 2,542 | -11% | 0 | 0 | — |
case-16 | pass→pass | 5,683 | 5,163 | -9% | 1 | 1 | 0% | 1,164 | 1,882 | +62% | 0 | 0 | — |
case-17 | pass→fail | 16,736 | 4,196 | -75% | 1 | 1 | 0% | 2,927 | 1,099 | -62% | 0 | 0 | — |
case-18 | fail→fail | 3,507 | 4,541 | +29% | 1 | 1 | 0% | 657 | 1,055 | +61% | 0 | 0 | — |
case-19 | fail→fail | 5,483 | 4,022 | -27% | 1 | 1 | 0% | 944 | 1,013 | +7% | 0 | 0 | — |
case-20 | pass→fail | 9,907 | 5,059 | -49% | 1 | 1 | 0% | 1,900 | 1,061 | -44% | 0 | 0 | — |
case-22 | pass→fail | 10,111 | 5,829 | -42% | 1 | 1 | 0% | 1,881 | 1,119 | -41% | 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 11 counted toward the lift figure. The other 11 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 -18 percentage points is the difference between those two pass rates over the 11 comparable cases. 7 cases got worse with the skill loaded, and they are included in that figure.
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.