CtrlK
BlogDocsLog inGet started
Tessl Logo

context-degradation

Language models exhibit predictable degradation patterns as context length increases. Understanding these patterns is essential for diagnosing failures and designing resilient systems.

35

Quality

31%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

SecuritybySnyk

Passed

No findings from the security scan

Fix and improve this skill with Tessl

tessl review fix ./plugins/AI-Agents-Safe-Coding-Skills-claude/skills/context-degradation/SKILL.md

The canonical home for this skill is context-degradation in administrakt0r/AI-Agents-Safe-Coding-Skills

SKILL.md
Quality
Evals
Security

Quality

Content

25%Weight 40%Scale 1-5

Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.

The body is a well-sectioned but verbose knowledge dump that explains familiar concepts at length and offers abstract rather than executable guidance, with all reference material inlined and one dangling reference. Splitting empirical detail into reference files and tightening prose would substantially improve it.

Suggestions

Move the empirical benchmarks table and model-specific thresholds into a separate reference file (e.g. references/benchmarks.md) and link to it from the body, or remove the dangling 'Degradation Patterns Reference' if no file exists.

Trim the conceptual re-explanation of lost-in-middle and attention sinks to the unique, actionable specifics Claude does not already know.

Convert the Four-Bucket and architectural-patterns sections into concrete, executable steps (e.g. exact compaction trigger thresholds, sample sub-agent invocation patterns) with validation checkpoints.

DimensionReasoningScore

Conciseness

The ~225-line body is noticeably verbose and re-explains concepts Claude already knows (lost-in-middle, BOS attention sinks, what a distractor is) alongside padded prose, fitting 'noticeably verbose; several unnecessary explanations or padded sections' rather than a 1 only because genuine empirical data is mixed in.

2 / 5

Actionability

Guidance is mostly high-level hints ('Place critical information at beginning or end', 'Use just-in-time context loading', the Four-Bucket names) with no concrete commands or executable steps, matching 'minimal concrete guidance; high-level hints but missing the specific steps to execute'.

2 / 5

Workflow Clarity

There is rough conceptual organization (pattern -> detection -> mitigation) and an advisory numbered Guidelines list, but no sequenced operational workflow with validation checkpoints, fitting 'rough sequence present but many gaps; steps poorly defined; validation absent'.

2 / 5

Progressive Disclosure

No bundle files exist, the single file inlines reference-grade material (the RULER findings, model-specific threshold table, and per-pattern deep dives) that belongs in separate references, and the only named reference ('Degradation Patterns Reference') has no corresponding file, matching 'content that clearly belongs in separate files is inlined'.

2 / 5

Total

8

/

20

Passed

Description

37%Weight 40%Scale 1-5

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 names a clear domain but relies on abstract actions and technical jargon with no explicit trigger guidance, leaving it indistinct from related context-engineering skills. Adding a 'Use when...' clause with natural trigger phrases would lift completeness and trigger-term quality.

Suggestions

Append a 'Use when...' clause naming natural triggers, e.g. 'Use when an agent's output degrades in long conversations, when investigating lost-in-middle effects, or when designing systems for large contexts.'

Replace abstract actions ('diagnosing failures', 'designing resilient systems') with concrete capabilities like 'identifies degradation patterns, locates lost-in-middle failures, and applies compaction or isolation mitigations.'

Add recognizable trigger keywords and synonyms users actually say ('long context', 'lost in the middle', 'context window', 'agent performance degradation') to improve distinctiveness.

DimensionReasoningScore

Specificity

The description names the domain ('predictable degradation patterns as context length increases') but the only actions are abstract ('diagnosing failures and designing resilient systems') with no concrete operations, matching the 'names the domain but actions are minimal or generic' anchor rather than the 1-2 concrete actions needed for a 3.

2 / 5

Completeness

It clearly states what the skill covers (understanding degradation patterns for diagnosis and design) but has no 'Use when...' or equivalent trigger guidance, so per the rubric cap completeness is held at 3 rather than rising to 4.

3 / 5

Trigger Term Quality

It contains one or two recognizable keywords ('context length', 'degradation') but is otherwise technical jargon ('resilient systems', 'diagnosing failures') with none of the natural phrases a user would say ('lost in middle', 'agent degrading in long conversation'), fitting the 'one or two generic keywords; missing the natural phrases' anchor.

2 / 5

Distinctiveness Conflict Risk

The context-degradation niche is somewhat specific but the abstract phrasing ('language models', 'designing resilient systems') could overlap with adjacent context-engineering skills and lacks distinct triggers, matching the 'somewhat specific but could still overlap' anchor.

3 / 5

Total

10

/

20

Passed

Validation

93%

Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.

Validation15 / 16 Passed

Validation for skill structure

CriteriaDescriptionResult

frontmatter_unknown_keys

Unknown frontmatter key(s) found; consider removing or moving to metadata

Warning

Total

15

/

16

Passed

Repository
administrakt0r/AI-Agents-Safe-Coding-Skills
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.