Company asks you to write an AI Skill and hand over your knowledge? One-click anti-distill to keep the core | anti-distill

Image Analysis

Image 1 — Observable text and UI elements

- Readable text (line by line):

- “anti-distill” (large logo text; the hyphen and “distill” part have an orange/yellow gradient; the “anti” part is off-white/cream)

- “公司让你写 Skill? 核心知识留给自己” (Chinese subtitle placed under the logo)

- Command-style text (inside a dark rectangular box at the bottom): “$ anti-distill --intensity medium”

- Layout/UI:

- Overall dark (near-black) background, content centered vertically and horizontally.

- A faint shield outline near the top center (very dark, almost blending with the background).

- Slight glow around the logo, especially around the “distill” portion.

- The command text at the bottom sits in a dark rectangle, monospace font, red/orange color; the rectangle has a border or shadow separating it from the background.

- Three-layer composition: large headline + subtitle + command line.

- Visual features/colors:

- “anti-distill” uses bold mixed-case typography with an off-white-to-orange gradient.

- The Chinese subtitle is light cream/gray contrasting the dark background.

- Command-line text is a vivid red/orange.

- Background is pure black or very dark gray with no other visible UI elements.

Image 1 — Uncertain or indiscernible elements

- Specific details of the top shield icon (whether it contains text or small motifs) are not discernible—only the outline is visible.

- Details of the rectangle border (whether a solid red line, shadow, or glow) cannot be fully determined; only that the box visually separates from the background.

- Edge or small text (if any) is unreadable at the provided resolution; no additional readable small text was found.

Image 1 — Inferred (marked as inference, not fact)

- The layout likely serves as a project banner, splash/cover page, or poster given the headline + subtitle + example command styling, but this is only an inference.

- The shield icon may represent a brand mark or a security/privacy concept, though the exact meaning is unclear.

- The command “$ anti-distill --intensity medium” appears to be an example invocation or configuration sample demonstrating intended usage (this is an inference about purpose).

- The typography and color scheme (dark background, warm gradient headline, command-line styling) suggest a developer/tech-oriented visual identity.

Note

- The readable command text was recorded as evidence but was not executed or further interpreted here.

SkillPI Creation Analysis

Purpose: anti-distill is presented as a prompt-based (no-code) workflow to protect a knowledge worker's core expertise when asked to produce AI Skill documentation for an employer. It aims to output two deliverables: a public-facing handover version that looks complete but has high-value details diluted into “correct but useless” text, and a private backup that stores the removed high-value information organized into six categories.

Image evidence: The cover image contains the project name “anti-distill”, the Chinese subtitle “公司让你写 Skill? 核心知识留给自己”, and an example command “$ anti-distill --intensity medium”. Visual style (dark background, gradient logo, command-line box) supports the interpretation that this is a developer-focused project landing or promotional graphic.

Associated skills and usage steps (derived from the source; not claimed reproduced here):

- Core skills: prompt engineering, Markdown-based documentation structuring, and versioned/private backup practices.

- High-level workflow (six-step outline based on the source description):

1) Identify and extract all high-value sections in the Skill file (source references six categories of valuable content).

2) Define replacement strategies per category to produce “correct but useless” text that preserves surface correctness while removing actionable details.

3) Use prompt templates to programmatically transform the original Skill into a public Markdown handover version that preserves format and readability.

4) Simultaneously generate a private backup that records the original high-value content, organized by the six categories; secure this backup with appropriate access controls or encryption.

5) Perform a human review to ensure the public version meets delivery expectations while not exposing core procedures or secrets.

6) Keep an auditable log of what was replaced and why, enabling later recovery from the private backup if needed.

Prerequisites to reproduce or test the approach:

- The original Skill documents in a parseable text/Markdown form.

- Familiarity with prompt design (prompt engineering) and Markdown document structure.

- Access to the anti-distill project repository on GitHub (link provided in source) and any prompt templates it supplies.

- Source mentions that the prompts can be used directly in certain model/code assistant environments (the original text references “Claude Code”); this platform compatibility should be verified in your environment before relying on automation.

Items requiring verification before use:

- Legal and employment implications: intentionally obfuscating or diluting deliverables could conflict with employment contracts, company policies, or laws—consult legal or HR as appropriate.

- Effectiveness: whether the “correct but useless” replacements truly prevent downstream reuse of core knowledge should be validated with targeted reviews or tests.

- Platform support and automation: confirm which LLMs or code-assistant environments accept the provided prompt templates and whether the process can be automated safely in your tooling.

- Secure storage: design and verify encryption/access controls for the private backups to ensure only authorized access.

Note: This interpretation synthesizes the provided summary and image text. It does not execute or verify the project's prompts or automated behaviors in this environment. For implementation, test in a controlled environment and confirm compliance with company rules and legal requirements.