contextrot
Priyanshu-byte-coder/contextrot
Contextrot is a local command-line tool that analyzes your coding agent's session transcripts to measure context degradation, calculate failure thresholds, and provide actionable optimization prescriptions. It runs entirely on your machine without requiring API keys or data uploads.
49 7MITUpdated 2026-07-25
86
GreatOverview
Contextrot solves the problem of declining coding agent performance as context windows fill up by analyzing historical session logs stored locally on your machine. By tracking five independent failure signals—such as edit failures, retry loops, re-reads, self-corrections, and tool errors—it correlates behavioral degradation with context fill percentages. The tool delivers an honest verdict, compares performance across models and projects, and integrates directly with your development environment via live status lines, hooks, and MCP servers.
Capabilities
- ▸Parses local JSONL session transcripts from coding agent CLIs
- ▸Calculates Wilson 95% confidence intervals and multi-signal failure metrics
- ▸Generates head-to-head project and agent performance comparisons
- ▸Produces interactive local HTML reports and exportable SVG badges
- ▸Provides real-time terminal status line updates and MCP server capabilities
Best for
Determine the exact context fill percentage where your coding agent starts failing, Compare failure rates and performance across different model families like Opus, Sonnet, and Haiku, Identify which project repositories or agent CLIs degrade the fastest, Integrate live context-health meters into terminal status bars, tmux, or Claude Code, Receive local HTML reports and shareable summary cards detailing agent efficiency
Works with
Claude CodeCodexGemini CLIOpenCode
Evoa Score breakdown= Σ (score × weight)
86Task usefulness20%88 → +17.6
実タスクにどれだけ役立つか（機能の豊富さ・用途の明確さ）。 — AIによるcapabilities/use-cases解析
Code quality15%95 → +14.3
実装・指示の品質。 — AIによるSKILL.md/README解析
Maintenance15%100 → +15.0
リポジトリがどれだけ活発に保守されているか。 — GitHub 最終push日時の新しさ
Documentation12%57 → +6.8
ドキュメントの充実度・分かりやすさ。 — README/独自要約の情報量
Security15%100 → +15.0
危険・不審な挙動が無いか。 — AIによるセキュリティレビュー
Originality10%90 → +9.0
ありふれたラッパーではない独自性。 — AIによる独自性判定
Popularity8%45 → +3.6
コミュニティの採用度。 — GitHub Stars/Forks（対数スケール）
Compatibility5%100 → +5.0
対応AIエージェントの広さ。 — AIによる対応エージェント判定
Weighted total86.3 / 100
ライセンス不明/制限あり（Red）のSkillは総合スコアに0.85倍の補正を適用します。 ランキングはこのScoreのみで決まり、広告で変わりません。 算出方法の詳細 →
Security considerations
The skill operates entirely locally by reading standard JSONL transcript files already stored on disk by agent CLIs. It requires no network access, API keys, or remote data uploads, maintaining complete user data privacy.
Categories
Summary and analysis are original content generated by AI Skills Rank. The skill's source text is not reproduced here — view it on the linked repository.