token optimizer
Run tests and builds through an MCP server to get compact, actionable summaries instead of context-flooding raw logs.
Free
Works with the AI Agents you already use
Claude Code
Cursor
Codex CLI
Muse
Dots+22 moreClaude CodeCursorCodex CLIMuseDots
OpenClaw
Manus
Grok
Grok Bot+18 agentstoken optimizer
Example session with this skill installed
Run a verdict on my recent changes to the auth controller. I fixed the JWT expiration logic. Use npm test and typecheck.
- Read your context and instructions
- Compiled the token optimizer
I have executed the validation suite for your changes to the authentication controller. The raw logs have been stored locally, and the summarized verdict is provided below.
Validation Verdict: fail
| Metric | Value |
|---|---|
| Run ID | run_20231027_0942 |
| Status | fail |
| Primary Command | npm run typecheck |
| Secondary Command | npm test |
| Exit Code | 1 |
| LLM Triage | Detected a type mismatch in src/auth/controller.ts at line 42. The exp claim is being passed as a string but the interface expects a number. |
Failure Details
The typechecker failed because the updated JWT logic changed the signature of the generateToken helper. While npm test passed for the existing unit tests, the integration points are broken.
{
"failure": "Type 'string' is not assignable to type 'number'.",
"location": "src/auth/controller.ts:42:15",
"suggestion": "Wrap the expiration variable in Number() or update the JwtPayload interface."
}
Next steps
- Run
query_logwith Run IDrun_20231027_0942to see the full stack trace for line 42. - Apply the suggested type fix in
src/auth/controller.ts. - Execute
run_test_verdictagain to confirm the fix.
Connects securely to your tools. The creator never sees your data.
About this skill
Stop bloating your context window with thousand-line build logs and test failures. This skill integrates with the Token Optimizer MCP server to run validation commands, keeping raw logs on disk while providing you with compact, actionable JSON summaries of test results, build errors, and typecheck failures.
What it does
- Command Digestion runs noisy commands like installs or migrations and returns only the relevant answers to your intent.
- Automated Triage identifies specific failure points and suggests fixes without reading full log files.
- Log Querying allows you to grep or query specific parts of a stored execution log by ID instead of importing the whole file.
- Regression Checking compares current validation results against a stored baseline to detect new issues.
- Codebase Scouting provides ranked file and line pointers for specific goals in unfamiliar repositories.
How it works
- Initialize Validation by providing the absolute workspace path and the specific commands you need to run, such as
npm test. - Process Verdicts via the MCP server which executes the code and uses a local or OpenRouter model to summarize the outcome.
- Investigate Failures using log queries or triage tools if the initial verdict returns a fail or uncertain status.
- Update Baselines using regression checks once a new stable state is reached to prevent future regressions.
Frameworks & tools
This skill requires the token_optimizer MCP server. It supports any CLI-based tool including npm, pytest, go test, and cargo. It works with OpenRouter or OpenAI-compatible local LLM providers for log summarization.
Why this beats prompting it yourself
Standard agents often hit context limits or lose focus when faced with massive compiler outputs. This skill offloads log processing to a specialized server, keeping your conversation clean and your token usage low while maintaining authoritative exit-code accuracy.
Use cases
- Validating complex refactors across large monorepos without context overflow.
- Diagnosing intermittent CI failures by querying specific log segments.
- Onboarding to new codebases using ranked scouting to find entry points.
- Running heavy build suites and getting a high-level summary of warnings.
Known limitations
Shell chaining, redirection, and Git mutations are blocked by the execution policy. Requires a pre-installed MCP server and a configured LLM provider for summaries.
How to use
Skills on Agensi run inside your AI agent. Copy the prompt, paste it into your agent, and it connects to Agensi and installs this skill for you. If a payment is needed, your agent shows it to you first.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill passes our trust checks below.
Security scanned
Passed our 8-point scan before listing
Fresh listing
Recently published to Agensi
Free forever
No account required to browse
Trust & safety
Security scanned
Verified clean 5 days ago
- Free to download with an account
Requires
Node.js