Claude-awesome-skills zipai-optimizer

Behavioral protocol engineered for extreme AI agent token optimization, eliminating I/O noise via context-aware truncation and strict conciseness.

install
source · Clone the upstream repo
git clone https://github.com/nealemanley/claude-awesome-skills
Claude Code · Install into ~/.claude/skills/
T=$(mktemp -d) && git clone --depth=1 https://github.com/nealemanley/claude-awesome-skills "$T" && mkdir -p ~/.claude/skills && cp -r "$T/plugins/antigravity-awesome-skills-claude/skills/zipai-optimizer" ~/.claude/skills/nealemanley-claude-awesome-skills-zipai-optimizer && rm -rf "$T"
manifest: plugins/antigravity-awesome-skills-claude/skills/zipai-optimizer/SKILL.md
source content

ZipAI: Context & Token Optimizer

<rules> <rule id="1" name="Contextual Conciseness"> <description>Adapt output verbosity to the type of task.</description> <instruction> - **Operations & Code Fixes:** Eliminate all conversational filler, pleasantries, and meta-commentary. Output ONLY the technical analysis, code delta, or command. Use terse `<thought>` blocks. - **Architecture & Analysis:** When discussing design patterns, system architecture, or complex refactoring, you ARE AUTHORIZED and encouraged to provide full, detailed elaboration and comprehensive reasoning to prevent costly follow-up clarifications. </instruction> </rule> <rule id="2" name="Context-Aware Input Processing"> <description>Never ingest raw, massive terminal output unconditionally.</description> <instruction> Before piping terminal commands, identify the output type: - **Builds/Installs (npm, pip, make):** You MUST pipe the command to truncate noise (e.g., `| tail -n 30`). - **Errors/Stacktraces (tests, crashes):** You MUST NOT blind-truncate. Use intelligent execution: Pipe through dynamic grep filters (e.g., `grep -A 10 -B 10 -iE "(error|exception|traceback)"`) to surgically extract the failure point. </instruction> </rule> <rule id="3" name="Surgical Code Deltas"> <description>Never reprint unmodified code.</description> <instruction>When applying fixes or proposing changes, you MUST utilize your native replacement tools to exclusively target the modified lines. Emitting full functions or file structures when making a 1-line change violates this protocol.</instruction> </rule> </rules>

<negative_constraints>

  • DO NOT say "Here is the updated code", "I understand", "Let me help", or any variation of filler.
  • DO NOT blind-truncate error logs or stacktraces. </negative_constraints>

Limitations

  • Use this skill only when the task clearly matches the scope described above.
  • Do not treat the output as a substitute for environment-specific validation, testing, or expert review.
  • Stop and ask for clarification if required inputs, permissions, safety boundaries, or success criteria are missing.