context-degradation
This skill should be used when the user asks to "diagnose context problems", "fix lost-in-middle issues", "debug agent failures", "understand context poisoning", or mentions context degradation, attention patterns, context clash, context confusion, or agent performance degradation. Provides patterns for recognizing and mitigating context failures.
84
80%
Does it follow best practices?
Impact
Pending
No eval scenarios have been run
Passed
No known issues
Optimize this skill with Tessl
npx tessl skill review --optimize ./skills/context-degradation/SKILL.mdQuality
Discovery
89%Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.
This description excels at trigger term coverage and completeness with a clear 'Use when...' clause containing numerous specific phrases. The main weakness is the vague capability statement - 'provides patterns for recognizing and mitigating' doesn't specify what concrete actions the skill enables. The description would benefit from listing specific diagnostic or remediation actions.
Suggestions
Replace 'Provides patterns for recognizing and mitigating context failures' with specific actions like 'Analyzes context windows for attention degradation, identifies poisoned segments, restructures prompts to avoid lost-in-middle effects'
Add concrete outputs or deliverables the skill produces (e.g., 'generates context health reports', 'recommends prompt restructuring strategies')
| Dimension | Reasoning | Score |
|---|---|---|
Specificity | Names the domain (context problems, agent failures) and mentions 'patterns for recognizing and mitigating context failures' but lacks specific concrete actions like 'analyze attention patterns', 'restructure prompts', or 'identify poisoned context segments'. | 2 / 3 |
Completeness | Explicitly answers both what ('Provides patterns for recognizing and mitigating context failures') and when (detailed 'Use when...' clause with specific trigger phrases and terms). | 3 / 3 |
Trigger Term Quality | Excellent coverage of natural trigger terms users would say: 'diagnose context problems', 'fix lost-in-middle issues', 'debug agent failures', 'context poisoning', 'context degradation', 'attention patterns', 'context clash', 'agent performance degradation'. | 3 / 3 |
Distinctiveness Conflict Risk | Highly distinctive niche focused on context-specific debugging for agents. Terms like 'lost-in-middle', 'context poisoning', 'attention patterns' are specialized and unlikely to conflict with general debugging or document skills. | 3 / 3 |
Total | 11 / 12 Passed |
Implementation
70%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
This is a well-structured conceptual skill that excels at organizing complex information about context degradation patterns with clear workflows and excellent progressive disclosure. However, it leans toward explanation over execution—providing frameworks and concepts rather than concrete, copy-paste-ready code or commands. The content would benefit from more actionable implementation examples.
Suggestions
Add executable Python/pseudocode for detecting degradation thresholds (e.g., a function that monitors token count and triggers compaction)
Include concrete code examples for implementing the Four-Bucket strategies, particularly 'Select' (relevance filtering) and 'Compress' (summarization patterns)
Trim the 'Core Concepts' section which largely duplicates content from 'Detailed Topics' - either consolidate or make Core Concepts a brief summary with forward references
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The skill is comprehensive but includes some verbose explanations that Claude would already understand (e.g., explaining what attention mechanics are, general descriptions of how models work). Some sections could be tightened, particularly the 'Core Concepts' which restates ideas that appear in detailed sections. | 2 / 3 |
Actionability | Provides conceptual guidance and frameworks (Four-Bucket Mitigation) but lacks executable code examples. The YAML examples are illustrative rather than copy-paste ready. Missing concrete detection scripts, threshold-checking code, or specific tool commands for implementing the strategies described. | 2 / 3 |
Workflow Clarity | Clear multi-step processes with explicit decision points. The Four-Bucket Mitigation Framework provides clear sequencing with triggers ('Use when context utilization exceeds 70%'). Detection-to-recovery workflows are well-defined with explicit checkpoints for each degradation pattern. | 3 / 3 |
Progressive Disclosure | Excellent structure with clear overview sections leading to detailed topics. References section provides well-signaled one-level-deep links with 'Read when' guidance. Content is appropriately split between main skill and referenced materials (patterns.md, related skills). | 3 / 3 |
Total | 10 / 12 Passed |
Validation
100%Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.
Validation — 11 / 11 Passed
Validation for skill structure
No warnings or errors.
- Commit
3ab8c94
- 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.