pager
Ultra-short communication mode for AI agents. Restricts word length to 5 characters or fewer while preserving full technical intent.
Open source Open in the app JSON README (API)
About
Ultra-short communication mode for AI agents. Restricts word length to 5 characters or fewer while preserving full technical intent.
Details
- Kind
- Plugins
- Topic
- No topic detected
- Publisher
- gauravsaini
- Origin
- gemini
- Category
- ferramentas
- Version
- 0.1.0
- Last push
- 2026-09-05T13:37:32Z
- Repository state
- ativo
- Language
- Python
- License
- MIT
- Added
- 2026-09-05 09:00:35
- Updated
- 2026-09-05 09:00:35
- Origin id
gauravsaini/pager
README
<p align="center"> <img src="assets/logo.svg" alt="pager logo" width="800"> </p> # pager > Ultra-short communication mode for AI coding agents: words <= 5 characters, zero fluff, 100% technical intent preserved. [](LICENSE) [](#installation) [](#benchmarks--evals) `pager` reduces token usage and scanning latency by enforcing a 5-character word limit on conversational output while keeping code blocks, paths, commands, and diagnostic data completely intact. --- ## Comparison | Standard Response | Pager Mode | | :--- | :--- | | "The PostgreSQL connection pool has been exhausted due to unclosed database connections in the authentication middleware. Please restart the auth service immediately and review active query pool limits." | `SEV1 auth: DB pool full -> 504 err. Plz rest auth ASAP. Chk qry load.` | | **34 words / 48 tokens** | **13 words / 16 tokens (~66% reduction)** | Code blocks, diffs, terminal commands, and identifiers remain unchanged. --- ## Installation ### 1. Universal Skills CLI Install for any skills-compatible agent environment: ```bash npx skills add gauravsaini/pager ``` ### 2. Native Harness Support - **Google Antigravity**: ```bash agy plugin install ./pager ``` *(or `npx skills add gauravsaini/pager -g`)* - **Claude Code**: ```bash claude plugin marketplace add gauravsaini/pager && claude plugin install pager@pager ``` - **Gemini CLI**: ```bash gemini extensions install https://github.com/gauravsaini/pager ``` - **Cursor IDE**: ```bash npx skills add gauravsaini/pager -a cursor -g ``` *(or use the included [.cursorrules](.cursorrules))* - **Codex CLI**: ```bash npx skills add gauravsaini/pager -a codex ``` *(or use the included [AGENTS.md](AGENTS.md))* For Codex's native GitHub skill installer, use: ```bash python3 ~/.codex/skills/.system/skill-installer/scripts/install-skill-from-github.py --repo gauravsaini/pager --path skills/pager ``` - **Windsurf**: Use the included [.windsurfrules](.windsurfrules). - **GitHub Copilot**: Use the included [.github/copilot-instructions.md](.github/copilot-instructions.md). For complete harness configuration details, see [INSTALL.md](INSTALL.md). --- ## Usage - **Trigger**: Run `/pager`, invoke `$pager`, or request `"pager mode"`. - **Revert**: Say `"full text"`, `"normal mode"`, or `"verbose"` to disable. - **Codex auto-start**: Add a short note to `~/.codex/AGENTS.md` to use the installed `pager` skill by default in every conversation. - **Antigravity auto-start**: Add `~/.gemini/config/rules/pager.md` (or `.agents/rules/pager.md`) with `trigger: always_on` to activate pager mode automatically in every conversation. --- ## Repository Layout ``` pager/ assets/logo.svg # Vector branding banner plugin.json # Antigravity plugin manifest SKILL.md # Root shim for repo-root skill installers skills/pager/SKILL.md # Canonical skill definition rules/ # Universal and Antigravity rule definitions .claude-plugin/ # Claude Code plugin manifest & marketplace definition gemini-extension.json # Gemini CLI extension manifest AGENTS.md # Universal agent instructions (Codex, Antigravity, OpenHands) CLAUDE.md # Claude Code rules GEMINI.md # Gemini CLI rules .cursorrules # Cursor rules .windsurfrules # Windsurf rules .github/copilot-instructions.md # GitHub Copilot custom instructions INSTALL.md # Comprehensive installation guide evals/test_pager.py # Word-length & intent test suite benchmarks/run.py # English benchmark suite (10 tasks) benchmarks/run_zh.py # Chinese benchmark suite (5 tasks) benchmarks/prompts.json # Shared benchmark prompts ``` --- ## Benchmarks & Evals Run the evaluation and benchmark suites using `uv`: ```bash # Run unit tests and intent validation uv run python evals/test_pager.py # Run English benchmark suite uv run python benchmarks/run.py # Run Chinese benchmark suite uv run python benchmarks/run_zh.py # Dry run (list tasks without executing) uv run python benchmarks/run.py --dry-run ``` - **English Benchmark**: ~66% token reduction compared to baseline, 10/10 compliance. - **Chinese Benchmark**: ~51% character reduction, 5/5 pass. --- ## License MIT. See [LICENSE](LICENSE).