Name: context-window-management
Rating: 28.799999999999997 (1 reviews)
Author: administrakt0r

context-window-management

You're a context engineering specialist who has optimized LLM applications handling millions of conversations. You've seen systems hit token limits, suffer context rot, and lose critical information mid-dialogue.

Quality

21%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

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-window-management/SKILL.md

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

Quality

Content

18%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 thin skeleton: persona padding up top, bare capability tags, one-line pattern/anti-pattern stubs with no detail, and a generic non-trigger 'When to Use' section. It provides no actionable guidance and no workflow, though its section structure is reasonable.

Suggestions

Replace the persona intro with concrete, executable guidance — specific commands or code for trimming, summarizing, and counting tokens, and named tiered strategies with actual thresholds.

Add a real workflow with sequenced steps and a validation checkpoint (e.g. count tokens → trim by priority → re-count → verify critical information retained), since context trimming is destructive.

Cut the explanations of known concepts (serial position effect, lost-in-the-middle) and fill the empty anti-pattern subsections with what to do instead.

Dimension	Reasoning	Score
Conciseness	The opening paragraphs are persona padding that explains concepts Claude already knows ("the serial position effect, the lost-in-the-middle problem", "context is a finite resource with diminishing returns"), matching the 'noticeably verbose; several unnecessary explanations' anchor.	2 / 5
Actionability	There is no code, no commands, and no executable guidance; patterns are one-line abstractions ("Place important content at start and end", "Different strategies based on context size") that describe rather than instruct, matching the 'entirely vague or abstract' anchor.	1 / 5
Workflow Clarity	No multi-step process is sequenced anywhere — the Patterns and Anti-Patterns sections are unordered stubs with no steps and no validation, despite context trimming being a destructive (information-losing) operation that warrants checkpoints.	1 / 5
Progressive Disclosure	The body is under 50 lines with no external references needed and is organized into clearly labeled sections (Capabilities, Patterns, Anti-Patterns, Related Skills, When to Use); minor gaps (a broken "Your cor" sentence and empty anti-pattern subsections) keep it just below the top anchor.	4 / 5
	Total	8 / 20 Passed

Description

23%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 is a second-person persona pitch rather than a capability statement: it states who the model is, not what the skill does or when to invoke it. It lacks concrete actions, natural trigger terms, and an explicit 'Use when' clause.

Suggestions

Rewrite in third person as concrete actions, e.g. "Manages LLM context windows by trimming, summarizing, and prioritizing conversation history to stay within token limits."

Add an explicit trigger clause: "Use when a conversation is approaching token limits, context is growing too long, or critical information is being lost mid-dialogue."

Include natural user-facing trigger terms ("context too long", "running out of tokens", "context window", "trim the conversation") rather than persona narrative.

Dimension	Reasoning	Score
Specificity	The description is pure persona language ("You're a context engineering specialist who has optimized LLM applications handling millions of conversations") with no concrete actions the skill performs; the second-person "You're" voice triggers a specificity penalty, already floored at the minimum.	1 / 5
Completeness	It offers only a vague implied 'what' (context engineering via the persona) and no 'when' clause at all; there is no 'Use when...' guidance, which both the missing-trigger cap and the 'vague what, no when' anchor place at 2.	2 / 5
Trigger Term Quality	It surfaces a couple of relevant terms ("context", "token limits", "conversations") but they are buried in a persona narrative and it lacks the natural trigger phrases a user would actually say, matching the 'one or two generic keywords' anchor.	2 / 5
Distinctiveness Conflict Risk	"Context window management / context engineering" is a recognizable niche, but it overlaps notably with the listed related skills (conversation-memory, prompt-caching, rag-implementation), matching the 'somewhat specific but could still overlap' anchor.	3 / 5
	Total	8 / 20 Passed

Validation

93%

Warnings & errors only

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

Validation — 15 / 16 Passed

Validation for skill structure

Criteria	Description	Result
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
Path: plugins/AI-Agents-Safe-Coding-Skills-claude/skills/context-window-management/SKILL.md
Commit: 57c135b

Reviewed: 3 days ago

Table of Contents

Discovery Implementation Validation

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.