Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    coreball context

    2

    Focus your agent on task-relevant code using CoreBall to rank files and symbols without hitting context limits.

    Free

    1 useSecurity scanned
    coreball context

    coreball context

    Example session with this skill installed

    Use CoreBall to find where the authentication middleware is implemented and how it handles expired JWT tokens. Limit to 2000 tokens.

    • Read your context and instructions
    • Compiled the coreball context

    I have analyzed the repository using CoreBall to trace the authentication flow and JWT handling. By ranking files against your task, I identified the core middleware and the specific validation logic.

    Repository Context: Authentication Trace

    File PathRelevance ScoreRole
    src/middleware/auth.ts0.98Primary entry point for request authentication.
    src/lib/jwt.ts0.92Contains verifyToken and expiration check logic.
    src/errors/UnauthorizedError.ts0.75Defines the error thrown when tokens are expired.
    src/config/auth.config.ts0.61Stores secret keys and TTL configurations.

    Implementation Details

    The authentication logic is centered in src/middleware/auth.ts. It extracts the Bearer token and passes it to the verifyToken function in src/lib/jwt.ts.

    // excerpt from src/lib/jwt.ts
    export const verifyToken = (token: string) => {
      try {
        return jwt.verify(token, process.env.JWT_SECRET);
      } catch (err) {
        if (err.name === 'TokenExpiredError') {
          throw new UnauthorizedError('Token has expired');
        }
        throw new UnauthorizedError('Invalid token');
      }
    };
    

    The middleware catches the UnauthorizedError and returns a 401 status code with a JSON payload indicating the expiration.

    Next steps

    1. Review src/lib/jwt.ts to see if the expiration grace period needs adjustment.
    2. Examine the UnauthorizedError class to ensure the client receives the correct error code.
    3. Check tests/auth.test.ts to verify existing coverage for expired tokens.

    Connects securely to your tools. The creator never sees your data.

    About this skill

    Large codebases create too much noise for LLMs to process efficiently. This skill uses CoreBall to isolate task-relevant files and symbols, preventing context window bloat and reducing token costs. It stops the agent from reading the entire repository by ranking files based on your specific coding task.

    What it does

    • Task-based ranking identifies exactly which files and symbols relate to a specific bug or feature request.
    • Context packaging generates focused excerpts and file lists within a defined token budget.
    • Dependency tracing follows imports and callers to map out implementation details across the codebase.
    • Symbol search locates specific signatures or names without manual grepping.
    • Overview generation provides a repository-wide index when the task scope is still undefined.

    How it works

    1. Repository Discovery locates the root and checks for existing agent instructions or configurations.
    2. Query Refinement converts your high-level request into a concrete CoreBall ranking task.
    3. Context Retrieval invokes the CoreBall MCP server or CLI to fetch ranked file excerpts.
    4. Iterative Analysis reads source files, traces imports, and expands the search if gaps are found.

    Frameworks & tools

    This skill requires the CoreBall CLI or a configured CoreBall MCP server. It works across any language supported by CoreBall, respecting .gitignore rules by default to ensure only relevant source code is analyzed. You need to clone coreball from here: https://github.com/francomano/coreball

    Why this beats prompting it yourself

    Standard prompts often cause agents to hallucinate file paths or ingest thousands of lines of irrelevant code. This skill enforces a strict ranking and budgeting workflow, ensuring the agent only acts on verified, high-significance context.

    Use cases

    • Map out an unfamiliar repository to understand a new feature's entry point.
    • Locate the root cause of a specific error message across multiple microservices.
    • Prepare a focused context package before initiating a complex refactor.
    • Audit specific symbol dependencies before changing a core API signature.

    Known limitations

    Ranking and excerpts are navigation aids, not proof of completeness. It may omit generated code, unsupported languages, or files excluded by .gitignore unless specific flags are used.

    Does not install CoreBall; requires existing CLI or MCP server. Omissions may occur with generated code or unsupported languages. Minimum token budget of 128 required.

    How to use

    1. 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. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before any payment.

    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

    1 use

    Times used to date

    Free forever

    No account required to browse

    Trust & safety

    Security scan

    Security scan passed 1 month ago

    • Free with an Agensi account

    Permissions

    Read Files
    Network Access
    Terminal / Shell
    Listed1 month ago

    What's inside

    Frequently Asked Questions

    This skill uses CoreBall's ranking algorithms to identify and extract only the code snippets relevant to your specific task, preventing the agent from wasting tokens on irrelevant files or hitting context window limits.

    The skill is compatible with any AI agent that supports MCP (Model Context Protocol) or can execute CLI commands, and it works across all programming languages supported by the CoreBall engine.

    You will need to have the CoreBall CLI installed or a CoreBall MCP server configured on your local machine or environment for the skill to interface with your codebase.

    Yes, the skill automatically respects your project's .gitignore rules and local configurations to ensure the agent only analyzes and ranks actual source code, ignoring build artifacts and dependencies.

    Unlike manual prompting, this skill uses a structured ranking and dependency-tracing workflow to verify file paths and symbol relationships before the agent processes them, significantly reducing hallucinations.

    You receive the specialized agent skill definition, including the logic for task-based ranking, context packaging, and iterative dependency tracing, along with documentation for connecting it to your CoreBall instance.

    New in Agents & Orchestration