Your whole second brain, in a fraction of the space.

    1

    Your whole second brain, in a fraction of the space.

    Free

    Security scanned

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    Your whole second brain, in a fraction of the space.

    Example session with this skill installed

    Run search mode: How has the counterparty's position on the 'Falcon' project changed since the initial proposal in January?

    • Read your context and instructions
    • Compiled the your whole second
    • Generated the document

    I have resolved the house term 'Falcon' and the holder 'Counterparty' via the registry. I am analyzing the delta log for the 'Contractual Stance' axis.

    Position & Delta Report: Project Falcon

    DateHolderAxisDirectionDelta TypeSource Pointer
    2024-01-15Apex CorpPricingBasisInitial Positiondocs/proposals/p1_jan.pdf
    2024-02-10Apex CorpPricingHardenedBasis-Strengtheneddocs/emails/re_pricing_feb.eml
    2024-03-05Apex CorpScopeAwayRephrasedocs/memos/scope_update.md
    2024-03-22Apex CorpPricingReversedPosition & Basisdocs/transcripts/call_march_22.txt

    Current Position
    The counterparty has moved from a fixed-price stance to a variable-rate model following the March 22nd call. While the February 10th update reinforced the initial pricing with new market data (Basis-Strengthened), the most recent movement represents a full reversal in both stated wording and underlying evidential support.

    Trend Roll-up
    The trajectory shows high oscillation in the Pricing axis over a 90-day period. The shift from 'Hardened' to 'Reversed' within 40 days suggests the previous stance was a procedural posture rather than a settled commitment.

    Next steps

    1. Drill down into call_march_22.txt to verify the exact wording of the rate-cap exceptions.
    2. Update the House Terms register if 'Falcon' referents have expanded to include the new sub-project.
    3. Run Build Mode on the incoming/ folder to capture the latest April correspondence.

    p1_jan.pdf

    PDF · document

    Generated

    Example file from a real run - the skill writes it into your workspace.

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

    About this skill

    Your whole second brain, in a fraction of the space.

    Shannon is fast, small second-brain infrastructure for a large personal document store. Instead of loading the whole store into the model — hours of reading, a disk full of duplicated chunks and vectors, and a context window spent on loading data rather than reasoning about it — you keep a small standing abstract map of the whole store: atlas, cluster summaries and cards with verbatim keys and pointers. The model reasons over the map, with answers coming from ~1,500 loaded tokens and the originals fetched only when exact wording matters. That saves time, saves disk space, and leaves the maximum of the model's context free for reasoning. It works because nothing is duplicated: the Shannon method maps the "surprise" — the delta in meaning — across your documents without ever copying the source text into a database, and a position & delta register tracks how stances, opinions and their evidential bases change over time.

    What it does

    • Lossless Key Mapping extracts identifiers, amounts, and deadlines verbatim while abstracting narrative content into plain language.
    • Contextual Resolution parses lay questions into event schemas and expands them through a domain-specific thesaurus of technical equivalents.
    • Delta Tracking maintains a register of "meaning in motion," recording how positions, stances, and evidential bases shift over time.
    • Mechanical Probing executes high-precision joins across clusters, keys, and entities rather than relying on fuzzy semantic impressions.
    • Trend Analysis identifies trajectories like steady drift or sudden reversals by reading the delta log as a time series.

    How it works

    1. Derive the Frame by analyzing your historical record to determine what counts as a meaningful change for your specific domain.
    2. Build the Map by carding sources into a three-level hierarchy: Atlas, Cluster Summaries, and Individual Cards.
    3. Execute Search by resolving house terms and fragments, expanding the query vocabulary, and probing the map via a dedicated script.
    4. Refresh Dynamically by detecting file-system changes and appending new cards or delta records without rebuilding the entire index.

    Frameworks & tools

    This is a domain-agnostic logic layer designed for file-system-based document stores. It uses a standalone map_query.py script for grep-based probing and maintains state through plain-text deployment packs containing registries, thesauruses, and delta logs.

    Why this beats prompting it yourself

    Prompting an LLM to search large folders results in hallucinations or high-latency "needle in a haystack" failures. This skill provides a formal relational structure for unstructured data, ensuring every answer is anchored to a specific source pointer and that conflicting information is flagged rather than smoothed over.

    Use cases

    • Track how a counterparty's stance has hardened or softened across months of correspondence.
    • Resolve lay terminology to technical procedural labels in specialized document sets.
    • Monitor a research corpus for "slow drift" trends that individual document reads miss.
    • Audit specific figures and deadlines across a fragmented store without duplicating sensitive data.

    Known limitations

    Requires a structured deployment pack for domain-specific vocabulary. Does not support vector databases or embeddings by design. Initial map build requires a local file-system or cloud-store root.

    Measured results

    First deployment: 628 sources, 7,167,837 bytes of text → a map of 854,378 bytes (11.9%).

    Licensing

    Source-available under the PolyForm Noncommercial License 1.0.0: free for noncommercial purposes (personal use, research, study, and charitable, educational, public research, public safety or health, environmental and government organisations). Any commercial use requires a paid commercial licence from the author in advance, with no exceptions — whether the product built on Shannon is open or closed, sold or given away, software or a service. See LICENSING.md in the package.

    No support for vector stores/embeddings. Requires a predefined deployment pack. Performance depends on the quality of the derived frame. Large initial builds may require significant token usage.

    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

    Free forever

    No account required to browse

    Trust & safety

    Security scan

    Security scan passed today

    • Free with an Agensi account

    Permissions

    Read Files
    Listed6 days ago
    Updatedtoday

    What's inside

    Frequently Asked Questions

    Unlike RAG which relies on 'fuzzy' vector similarity, this skill uses the Shannon method to map 'meaning deltas.' It prevents information redundancy by tracking how data evolves over time while keeping the source text in its original location, eliminating the hallucination risks common in traditional vector searches.

    You receive the core logic layer, the `map_query.py` execution script, and the template structures for the Atlas, Cluster, and Card hierarchy. It also includes a deployment pack framework to help you build your domain-specific thesaurus and registries.

    This skill is a domain-agnostic logic layer that integrates with any AI agent capable of executing Python scripts and accessing a local or cloud-based file system. It is designed to sit between your agent and your document store as a high-precision retrieval engine.

    The initial build requires a one-time scan of your document root to 'Derive the Frame.' After that, the system uses dynamic refreshing to detect file-system changes, appending only new 'delta' records or cards without needing to re-index the entire corpus.

    Because this skill functions as a pointer-based map rather than a database, it never copies or stores your sensitive source text. It only holds metadata and 'verbatim keys,' meaning your actual data remains secure and untouched in its original environment.

    Source-available under the PolyForm Noncommercial License 1.0.0: free for noncommercial purposes (personal use, research, study, and charitable, educational, public research, public safety or health, environmental and government organisations). Any commercial use requires a paid commercial licence from the author in advance, with no exceptions — whether the product built on Shannon is open or closed, sold or given away, software or a service. See LICENSING.md in the package.

    New in Productivity