More screenshots

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    Context Rot Forensics

    1

    Separate position loss from fill saturation in failing long-context AI sessions.

    $19.99

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    Context Rot Forensics

    Context Rot Forensics

    Example session with this skill installed

    Diagnose this 40-turn coding session. The database schema appeared near the beginning, the session is now about 75% full, and later output stopped respecting that schema. Evidence E3 confirms the schema was still present but missed. Distinguish position effects from fill saturation, propose only justified repair moves, and define an A/B rerun with pass criteria.

    • Read your context and instructions
    • Compiled the context rot forensics

    Context Rot Diagnosis

    Verdict: FILL SATURATION
    Estimated fill: 75%
    Missed item: db_schema
    Observed zone: HEAD
    Depth: approximately 15% from the start
    Evidence cited: E3

    The critical schema was not buried in the BODY zone; it remained in the HEAD. That weakens a pure position-rot explanation. The stronger supported diagnosis is saturation at high fill, with the missed schema becoming harder to use reliably as the session accumulated turns. This is an evidence-bounded classification, not a claim about hidden model internals.

    Repair moves

    O5 — Applied inference: create a fresh state summary that pulls the schema and current decisions into a compact TAIL position.
    2.

    O7 — Applied inference: reset the session with a structured handoff at roughly 75% fill instead of continuing to accumulate context.

    A/B placement test

    • A: rerun the next task with the original 40-turn context unchanged.
    • B: rerun the same task with the schema and live decisions in a structured TAIL handoff.
    • Keep model, prompt, temperature, and evaluation criteria constant.
    • Pass condition: B follows every required schema field while A reproduces at least one omission.

    If both variants pass or both fail, the repair is not proven and the verdict remains unresolved for that task.

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

    About this skill

    Diagnose whether a long-context failure is position rot, fill saturation, a mixed case, or not context rot. Context Rot Forensics maps critical information across HEAD, BODY, and TAIL zones, estimates context fill, cites the evidence behind the missed item, distinguishes observation from inference, and proposes a layout repair only when the diagnosis supports one.

    The deliverable includes a scope-limited verdict, missed-item map, evidence ledger, repair moves, and an A/B placement test with exact pass criteria. The package includes the full SKILL.md workflow, CROT protocol, structured sample, deterministic context-map checker, tests, and a six-page PDF buyer guide.

    It does not increase a model's context window, guarantee recall, summarize documents, or claim a repair worked without the supplied A/B result.

    Does not increase a model's context window, guarantee recall, summarize documents, inspect hidden attention states, or prove a repair successful without the supplied A/B result. The verdict applies only to the session evidence provided.

    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 2 months ago

    Permissions

    No special permissions declared or detected

    Listed2 months ago

    What's inside

    Frequently Asked Questions

    This skill provides a systematic diagnostic framework to determine if an AI is failing due to the specific placement of information (Position Rot) versus the sheer volume of data (Fill Saturation). It helps you stop guessing why a long-context prompt failed and provides a structured 'repair move' to fix the document layout.

    This is a platform-agnostic diagnostic protocol that works with any LLM or AI agent experiencing performance degradation in long-context windows. You can apply the CROT protocol to workflows built in LangChain, AutoGPT, or custom enterprise agent frameworks.

    You receive the full SKILL.md workflow, the proprietary CROT protocol, a deterministic context-map checker, and a comprehensive six-page buyer guide. It also includes an A/B placement test framework to objectively verify if a diagnosis and subsequent repair actually improved recall.

    No, this skill is a diagnostic and optimization tool rather than a performance booster. It identifies why a model is ignoring data within its existing limits and provides a map to reorganize that data for maximum visibility, but it cannot expand the hardware-defined context window.

    Setup involves integrating the provided SKILL.md into your existing agent's system prompt or developer workflow. The included 'context-map checker' is a standalone utility that can be immediately used to audit your current long-context sessions.

    New in LLM Ops