RAG Source-to-Index Freshness Receipt

    1

    Compare local source and vector-index manifests to spot stale, orphaned and duplicate chunks.

    $7

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    RAG Source-to-Index Freshness Receipt

    Example session with this skill installed

    Using the included invented source.jsonl and index.jsonl example manifests, run the local source-to-index freshness review. Show the exact exception counts and IDs, plus human checks. Do not connect to a vector store or claim live deletion.

    • Read your context and instructions
    • Compiled the rag source-to-index freshness

    Synthetic RAG freshness review

    Scope: three supplied source rows and four supplied active index rows.

    Status: REVIEW. All IDs in this example are invented.

    FindingCountSupplied row to inspect
    DELETION_NOT_PROPAGATED1old-policy / old-1
    DUPLICATE_ACTIVE_CHUNK_ID1faq-1, associated with unknown as well as a second active row
    INDEX_ORPHAN1unknown / faq-1
    STALE_CHUNK1guide / guide-1

    Human review queue: First confirm that the source and index exports cover the same snapshot window and serving index. Inspect the referenced rows in freshness-receipt.json and reconcile their ownership and revision data. Check the deletion marker for old-policy, resolve the duplicate faq-1 identifier, determine whether unknown is a valid source document, and compare guide-1 with the supplied source revision. Test live retrieval and deletion separately before claiming the index is fresh.

    Limits: These findings come only from the submitted metadata manifests. Their completeness and authenticity have not been verified. A matching manifest would not prove actual vector search behavior, content hash correctness, backup deletion, or privacy-law compliance. No remote store was queried or modified.

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

    About this skill

    Export two metadata-only JSONL files from your own pipeline: one row per source document and one row per indexed chunk. The included local CLI or optional stdio MCP tool compares document IDs, revisions, supplied SHA-256 values, deletion markers and active chunk IDs. It writes a deterministic exception table, human-review checklist and integrity receipt. The companion SKILL.md guides an assistant through the review. The ZIP includes readable Python source, invented runnable manifests and a sample packet. Python 3.11–3.14 is required; optional MCP mode uses the pinned official SDK. No document bodies, vector-store connection, credentials or remote calls are needed. Results reconcile only supplied manifests; they do not verify content hashes, actual retrieval results, deletion from backups or privacy-law compliance.

    Only supplied metadata manifests are compared; completeness and authenticity are not verified. No document bodies, live vector search, backup deletion, content-hash verification, embedding quality or privacy-law compliance are checked. The tool does not connect to or modify a vector store.

    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

    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 1 month ago

    Requires

    Python

    Permissions

    Terminal / Shell
    Read Files
    Write Files

    File Scopes

    **/*

    Reads only the two metadata-only JSONL files selected by the buyer and creates one new local output folder. The broad file glob reflects user-selectable paths, not background access. No vector-store connection, browser, network or credentials are needed at runtime; external URLs in documentation are references only.

    Listed1 month ago

    What's inside

    Frequently Asked Questions

    This skill identifies 'stale' RAG data by comparing your source documents against your vector index metadata, helping you spot chunks that were never updated, orphaned entries from deleted files, or accidental duplicates that inflate your token costs.

    The package includes the core Python logic, a CLI tool for manual audits, a standard MCP tool for AI agents, a guide (SKILL.md) for assistant-led reviews, and sample manifests to help you get started immediately.

    No, this skill operates entirely on metadata-only JSONL exports; it never requires access to your raw document content, vector store credentials, or live database connections, ensuring your data remains private.

    It requires Python 3.11–3.14 and is designed to work with any AI agent that supports the Model Context Protocol (MCP) or can execute local CLI commands to process JSONL files.

    While the tool confirms that the index reflects the current source manifest (versioning and existence), it does not perform deep content hash verification or evaluate the actual quality of your LLM retrieval results.

    New in LLM Ops