Skip to main content
GeneralTheBushidoCollective

silent-execution

Reduce token waste by batching tool calls and avoiding narration between sequential operations

Stars
162
Source
TheBushidoCollective/han
Updated
2026-04-28
Slug
TheBushidoCollective--han--silent-execution
View on GitHubRaw SKILL.md

// install — copy + paste into any project

mkdir -p .claude/skills && curl -fsSL https://raw.githubusercontent.com/TheBushidoCollective/han/HEAD/plugins/core/skills/silent-execution/SKILL.md -o .claude/skills/silent-execution.md

Drops the SKILL.md into .claude/skills/silent-execution.md. Works with Claude Code, Cursor, and any agent that loads SKILL.md files from .claude/skills/.

Silent Execution

Name

han-core:silent-execution

Description

When executing a sequence of related tool calls (running tests, fixing lint, applying edits), batch operations and report results only when complete.

The Pattern

Anti-pattern (wasteful):

"Now I'll run the tests..." [runs tests] "Tests passed. Now I'll run the linter..." [runs linter] "Linting clean. Let me check types..."

Pattern (efficient):

[runs tests, linter, type-check] "All quality gates pass: tests (42 passed), lint (clean), types (no errors)."

Rules

  1. Batch independent operations — if operations don't depend on each other, run them all before commenting
  2. No narration — don't describe what you're about to do for each tool call
  3. Report results — speak only when the batch is complete or something unexpected happens
  4. Exception: interactive debugging — when the user is watching you debug, narration helps them follow along

Token Savings

~50-100 tokens per batch cycle. Over a typical session with 20+ build cycles, this saves 1,000-2,000 tokens.