silent-execution

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

Stars 129

Forks 14

Install this agent skill to your Project

npx add-skill https://github.com/TheBushidoCollective/han/tree/main/plugins/core/skills/silent-execution

SKILL.md

Silent Execution

Name

han-core:silent-execution

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

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)."

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

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