distill-shield
Distill Shield：为个人移交资料包生成 Canary 与可选加固策略，提高未授权蒸馏成本；仅适用于有权处置的数据。
52
60%
Does it follow best practices?
Impact
—
No eval scenarios have been run
Passed
No known issues
Optimize this skill with Tessl
npx tessl skill review --optimize ./distill-shield-skill/SKILL.mdQuality
Discovery
57%Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.
The description identifies a clear and distinctive niche—generating anti-distillation protections for handover data packages—which makes it unlikely to conflict with other skills. However, it lacks an explicit 'Use when...' trigger clause and relies on somewhat technical jargon that may not match natural user language, limiting its effectiveness for skill selection in a large pool.
Suggestions
Add an explicit 'Use when...' clause with natural trigger phrases, e.g., '当用户需要保护移交资料、防止未授权蒸馏、添加 Canary 水印或生成数据加固方案时使用'.
Include more natural language variations and synonyms users might say, such as '数据保护', '水印', '防泄露', '知识产权保护', to improve trigger term coverage.
| Dimension | Reasoning | Score |
|---|---|---|
Specificity | The description names a specific domain (distillation protection for personal handover packages) and mentions concrete outputs like 'Canary' and 'hardening strategies', but doesn't list multiple detailed actions beyond generation of these artifacts. | 2 / 3 |
Completeness | The 'what' is reasonably clear (generate Canary and optional hardening strategies for handover packages), but there is no explicit 'Use when...' clause. The phrase '仅适用于有权处置的数据' is a constraint/caveat rather than a trigger condition, so the 'when' is only implied. | 2 / 3 |
Trigger Term Quality | Includes some relevant terms like 'Canary', '蒸馏' (distillation), '移交资料包' (handover package), and '加固策略' (hardening strategy), but these are fairly specialized/technical terms. Common user-facing trigger terms or natural language variations are limited. | 2 / 3 |
Distinctiveness Conflict Risk | This is a very niche skill focused specifically on anti-distillation protection with Canary tokens for handover packages. It is highly unlikely to conflict with other skills due to its specialized domain. | 3 / 3 |
Total | 9 / 12 Passed |
Implementation
62%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
This skill provides a well-structured phased workflow for generating anti-distillation canaries with a clear ethical framing and validation checklist. However, it leans more descriptive than actionable in the early phases (threat model, strategy selection) and relies heavily on an external script without showing concrete output examples or customization options. The strategy comparison table is a strength, but the skill could be more concise by trimming the preamble and language instructions.
Suggestions
Add an example of what `shield_gen.py` actually outputs (e.g., sample CANARY.txt content) so Claude knows what to expect and can verify correctness.
Make Phases 0-2 more actionable by providing specific questions to ask the user or decision trees rather than just listing considerations.
Move the '复制给 AI' preamble section to the end or remove it—it's meta-information that doesn't help Claude execute the skill.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Generally efficient but includes some unnecessary sections like the language instruction and the 'copy to AI' preamble that add overhead. The threat model and strategy table are useful but the overall framing could be tighter. | 2 / 3 |
Actionability | Provides concrete bash commands for running the generation script and a useful strategy comparison table, but the actual shield generation process relies entirely on an external script without showing what it produces or how to customize output. Phases 0-2 are descriptive rather than instructive. | 2 / 3 |
Workflow Clarity | Clear phased workflow (Phase 0-5) with logical sequencing from threat modeling through generation to self-check. Phase 4 provides an explicit validation checklist, and Phase 5 addresses communication to recipients. The checklist in Phase 4 serves as a verification checkpoint for this type of operation. | 3 / 3 |
Progressive Disclosure | References external files like `FOR_AI.md` and the `kit/shield_gen.py` script, but doesn't clearly signal what additional documentation exists or link to detailed references for strategies. The content is reasonably organized but the 'copy to AI' section at the top is awkwardly placed and the relationship to other files could be clearer. | 2 / 3 |
Total | 9 / 12 Passed |
Validation
81%Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.
Validation — 9 / 11 Passed
Validation for skill structure
| Criteria | Description | Result |
|---|---|---|
metadata_version | 'metadata.version' is missing | Warning |
metadata_field | 'metadata' should map string keys to string values | Warning |
Total | 9 / 11 Passed | |
- Repository
- agenmod/immortal-skill
- Commit
cdab91b
- Reviewed
Table of Contents
Is this your skill?
If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.