無料GitHub で公開

context-mode

MANDATORY routing rules for context-mode. Invoke whenever you analyze, count, filter, compare, search, parse, or transform data; fetch a URL; or run a data-heavy command — so raw bytes stay out of the context window.

mksglu/context-mode

インストール方法を見る

含まれるファイル（1）

SKILL.md1.7 KB

SKILL.md（原文）

インストールする前に、エージェントに与えられる指示の中身を確認できます。

context-mode — MANDATORY routing rules

context-mode MCP tools are available in GitHub Copilot CLI. These rules protect the context window from flooding — one unrouted command can dump tens of KB into the conversation. Follow them strictly.

Think in Code — MANDATORY

Analyze / count / filter / compare / search / parse / transform data: write code via ctx_execute(language, code) and console.log() only the answer. Do NOT read raw data into context. PROGRAM the analysis, do not COMPUTE it by reading. One script replaces ten tool calls.

BLOCKED — do NOT use

curl / wget — dumps raw HTTP into context. Use ctx_fetch_and_index(url, source) then ctx_search(queries).
Inline HTTP (node -e "fetch(...)", python -c "requests.get(...)") — use ctx_execute(language, code); only stdout enters context.
Reading large files to analyze — use ctx_execute_file(path, language, code).

Tool selection

GATHER: ctx_batch_execute(commands, queries) — runs all commands, auto-indexes, returns search. One call replaces dozens.
PROCESS: ctx_execute / ctx_execute_file — sandbox; only stdout enters context.
WEB: ctx_fetch_and_index(url, source) then ctx_search(queries) — raw HTML never enters context.
SEARCH: ctx_search(queries: [...]) — all questions in one call.

Write artifacts to FILES; return a path + one-line description, not inline dumps.

レビュー

まだレビューはありません。使ってみた感想をお寄せください。

ログインしてレビューを書く

同じリポジトリのスキル

概要と使いどころ

context-mode

無料

Use context-mode tools (ctx_execute, ctx_execute_file) instead of Bash/cat when processing large outputs. Triggers: "analyze logs", "summarize output", "process data", "parse JSON", "filter results", "extract errors", "check build output", "analyze dependencies", "process API response", "large file analysis", "page snapshot", "browser snapshot", "DOM structure", "inspect page", "accessibility tree", "Playwright snapshot", "run tests", "test output", "coverage report", "git log", "recent commits", "diff between branches", "list containers", "pod status", "disk usage", "fetch docs", "API reference", "index documentation", "call API", "check response", "query results", "find TODOs", "count lines", "codebase statistics", "security audit", "outdated packages", "dependency tree", "cloud resources", "CI/CD output". Also triggers on ANY MCP tool output that may exceed 20 lines. Subagent routing is handled automatically via PreToolUse hook.

日本語の概要は準備中です。原文の説明を表示しています。