Tool Release

HarnessTrim: a deterministic, benchmarked token-economy layer across Claude Code, Codex & OpenCode

HarnessTrim is a new tool that provides a deterministic, benchmarked token-economy layer across coding agent harnesses like Claude Code, Codex, and OpenCode. It uses idempotent reducers to cut token waste from noisy tool output, model verbosity, thinking tokens, and instruction files, without involving an LLM in the reduction path. The tool coordinates existing single-channel solutions (e.g., Caveman, RTK) behind a unified cross-harness policy, and is cache-aware to avoid touching cacheable prompts.

1 engagement·1 source·Sun, Jul 12, 2026, 10:04 AM

Coding agents leak tokens across several channels: noisy tool output (test logs, git diff, build spam), model verbosity, thinking tokens, and always-loaded instruction files. Existing tools each address only one channel (Caveman → output style, RTK → shell output). HarnessTrim coordinates them behind one cross-harness policy, using the hook/skill primitives the harnesses already expose. Key design choices: deterministic + idempotent reducers (same input → byte-identical output, no LLM in the reduction path), and cache-aware (reducers never touch the cacheable prompt). The tool is benchmarked, though specific benchmark numbers are not provided in the post.

Entities

Codex(tool)Claude Code(tool)OpenCode(tool)HarnessTrim(tool)Caveman(tool)RTK(tool)

ProductSat, Jul 11, 2026, 06:38 PM

ContextOps: open-source tool to audit and optimize LLM prompt context

ContextOps is an open-source tool that analyzes LLM prompts to detect token waste such as duplicated retrieval chunks, bloated system prompts, oversized conversation history, and repeated tool outputs. It helps developers reduce costs and improve model consistency by auditing what goes into the prompt before inference.

2 engagement·1 source·reddit

Tool ReleaseSat, Jul 11, 2026, 01:33 PM

RDXmin tool adds terse persona and context-saving compressor for Claude

A new tool called RDXmin extends Claude's output terseness with a terse persona system (three levels: lite, full, ultra) and a tool-output compressor that shrinks tool results before they enter context, reducing token usage. The compressor strips ANSI codes, keeps head/tail and error lines, and deduplicates, running behind a safety allowlist to avoid affecting Read/Edit/Write operations.

4 engagement·1 source·reddit

ProductSun, Jul 12, 2026, 12:47 PM

Coder: CLI tool to delegate coding tasks to background LLM agents

Coder is a CLI/plugin that lets developers dispatch coding tasks to background agents powered by Claude CLI or Codex CLI. It keeps the main session context clean and distributes workload across existing subscriptions. Built entirely with Claude Code, it is free and open source.

4 engagement·1 source·reddit

ProductSun, Jul 12, 2026, 09:53 AM

CLI to analyze Claude Code session token costs

Tokenbill is an open-source CLI tool that parses local Claude Code session logs and estimates token costs. It runs entirely offline with no network calls, helping developers understand and optimize their Claude Code usage costs.

2 engagement·1 source·reddit

ProductSat, Jul 11, 2026, 07:51 AM

Snitch: deterministic claim verifier for AI agent transcripts

Snitch is a tool that watches AI agent transcript files (Cursor, Claude Code, Codex, Pi, OpenCode) and verifies claims made in prose against actual evidence like tool calls, shell output, filesystem changes, git history, and session context. It uses deterministic regex patterns to extract claims and cross-references them, flagging inconsistencies. It helps developers trust their coding agents by catching when an agent's description doesn't match reality.

3 engagement·1 source·reddit