Works with the AI Agents you already use
Token-Plus
Reduce API costs and context bloat by enforcing targeted file reading and quiet command execution.
$15
Token-Plus
Example session with this skill installed
The session is getting expensive. Audit my current context for waste and give me a brief on this project so I don't have to keep re-reading the source files.
- Read your context and instructions
- Compiled the token-plus
I have audited the current session context and identified several sources of token waste, including a heavily populated node_modules listing and redundant reads of the main configuration file. I have also prepared a compact project brief to minimize future file scans.
Context Audit & Project Brief
| Category | Finding | Optimization Action |
|---|---|---|
| Waste Found | package-lock.json fully loaded in context (8,000+ lines). | Ignored in future turns; use npm list --depth=0 instead. |
| Redundancy | src/config.ts read 3 times in the last 10 turns. | Cached key definitions in session notes. |
| Noise | Full git log output (200 lines) occupies active context. | Switched to git log --oneline -n 10. |
Project Onboarding Notes
- Core Purpose: TypeScript-based CLI tool for repository token optimization.
- Key Entry Point:
src/index.ts:1-50handles command routing. - Main Logic:
src/core/optimizer.tscontains the logic for log triaging and search filtering. - Dependencies: Uses
ripgrepfor fast searching andsedfor line-range extraction. - Exclusions: Ignore
dist/,build/, and any.mapfiles during exploration.
Next steps
- I will clear the
package-lock.jsonfrom the current buffer to save ~24,000 tokens per turn. - I will use
grep -E "FAIL|Error"for the next test run to avoid loading the full log. - Refer to these onboarding notes instead of re-reading
src/index.ts.
Connects securely to your tools. The creator never sees your data.
About this skill
Long coding sessions and large codebases quickly bloat context windows, leading to high latency and soaring API costs. This skill provides a systematic framework for navigating, reading, and auditing repositories without wasting tokens on irrelevant data.
What it does
- Targeted Reading Locates specific definitions using narrow patterns before reading minimal line ranges in large files.
- Output Triage Filters massive logs and test outputs to extract only failures, errors, and summaries.
- Command Optimization Uses quiet flags and filtered listings for git, builds, and package managers to prevent terminal noise.
- Context Hygiene Tracks findings with compact notes to prevent redundant file reads and unnecessary re-scanning.
- Token Auditing Analyzes session history and loaded files to identify sources of waste and optimize future turns.
How it works
- Locate Execute a filename or match-count search to pinpoint the target without reading file contents.
- Read Range Access only 40 to 120 lines around a match instead of loading the entire file.
- Filter Output Use POSIX tools like
grep,sed,head, andtailto prune command results. - Audit Review the current session context to find and remove "always-loaded" waste like lockfiles or minified assets.
Frameworks & tools
This skill is compatible with any AI agent that supports SKILL.md instructions, including Claude Code, Cursor, and GitHub Copilot. It relies on standard POSIX shell utilities like git, grep, rg, sed, head, and tail.
Why this beats prompting it yourself
Standard prompts often cause agents to ingest full files or massive log outputs by default, which can triple session costs. This skill enforces a "locate-first" behavior and provides specific command flags that agents do not typically use, ensuring context remains lean without manual intervention.
Use cases
- Onboarding to a massive codebase without hitting context limits immediately.
- Debugging failing CI/CD logs that exceed thousands of lines of output.
- Running performance-intensive tasks like deep git history searches or bulk installs.
- Managing long-running agent sessions where cost and latency are primary concerns.
Known limitations
Does not optimize destructive commands, git status during active merges, or searches intended to prove non-existence. Secrets and .env files are always excluded from lean reading for safety.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for LLM Ops.
Does not optimize destructive commands, git status during merges, or exhaustive searches proving non-existence. Shell tips require a POSIX environment with git and grep/rg.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 11 days ago
- Software
- jest
Node.js
Git
Permissions
File Scopes
No external accounts or services needed. This skill only gives Claude instructions for using fewer tokens.