Keep the source within reach
Carry provenance with the claim instead of relying on an answer with no path back to its inputs.
Smart tool
Knowledge
Follow a claim back to the evidence.
For people connecting notes, documents and research. Knowledge turns supplied material into an inspectable graph, with sources and time attached, so you can trace relationships, question an apparent contradiction and see what changed.
Claim: the export supports the selected format Source A: specification, revision one Source B: compatibility note, revision two Relationship: candidate contradiction — needs review Inspect: source passages, validity windows, connecting path Decision: retain both claims until the actual export is checked Share: self-contained graph viewer
Fictional graph investigation. The claim and notes are illustrative; no knowledge-correctness result is implied.
Explore a specific question
Use paths, evidence, gaps and timelines to navigate relationships rather than rereading an entire collection.
Keep a history of changing claims
Close a relation’s validity window and append its replacement instead of erasing the earlier view.
The film
Keep the quote. Keep the date.
In 70 seconds, Mags the magpie grabs claims and leaves their sources behind, so a help page and a newer note disagree about when online orders close. Knowledge reads the notes into a map where every link keeps its quote, source and date: you get a timeline instead of one answer, a flag on the contradiction for you to decide, and a question back when you ask something vague. Captions are on the player; a transcript is below.
Read the transcript
Meet Mags. Mags collects shiny claims.
She grabs the claim, and leaves the source behind.
Months later, a customer orders at seven. The help page says too late. The newest note says eight.
Knowledge puts the source back on the claim.
Your agent reads the notes into a map. Every link keeps its quote, its source and its date.
Ask when orders close. You get a timeline: six until May twentieth, then eight.
The help page gets a flag. A possible contradiction, for you to decide.
Ask something vague, and it asks you back.
A link is a lead, not proof. Read the quote.
On screen: Keep the quote. Keep the date.
Knowledge. Follow a claim back to the evidence.
The bakery, its notes and the map are a fictional example. The film shows how the tool behaves; it does not report results.
See the work take shape.
Fictional graph investigation. The claim and notes are illustrative; no knowledge-correctness result is implied.
Choose the graph and input
Use structured additions or deterministic cooccurrence ingestion without a model. Optional extraction needs a configured provider.
Inspect the relationship
Follow a path or inspect its evidence, treating cooccurrence and candidate contradictions as signals to review.
Share an inspectable view
Export the graph or generate its self-contained offline viewer with the sources available for inspection.
Fictional example
Investigate two conflicting notes
The brief Use Knowledge to show how two fictional project notes support or challenge the same claim. Start with deterministic ingestion or structured additions. Trace each relationship to its source and distinguish an asserted fact from a candidate contradiction. Do not configure an external provider.
Before
Two notes disagree, but a keyword search presents each in isolation.
After applying the method
A temporal path and source-linked evidence expose the disagreement for review.
Relationships and provenance make the conflict inspectable.
A clear job.
A useful stopping point.
Ingest, connect, trace, challenge or explore knowledge from supplied material.
Use it when
- You need to trace a relationship across supplied documents.
- A claim changed over time and the earlier state matters.
- A team needs an offline view of a graph and its evidence.
Choose another route when
- Treating word cooccurrence as proof of a semantic relationship.
- Claiming that a structural graph gap proves a scientific research gap.
Build
Sources need an inspectable graph.
Nodes and edges with source, method and temporal provenance.
Investigate
A relationship, contradiction or gap needs examination.
Paths, evidence, timeline and qualified gaps.
Explore
A person needs to inspect or reuse the graph.
A self-contained viewer or portable graph export.
Read the canonical method
Knowledge
Use the retained knowledge CLI or knowledge.lib. Run knowledge --help and knowledge <capability> --help for exact arguments. Its deterministic core uses Python and SQLite; model-backed extraction, reconciliation and answers are optional capabilities.
Build: choose the graph explicitly, preserve source identity and distinguish cooccurrence from an asserted semantic relationship. knowledge ingest --method cooccurrence provides a no-model path; use add for already structured evidence. Inspect extraction provenance before treating edges as facts.
Investigate: use path, evidence, contradictions, gaps and timeline to test a specific question. Read references/graph.v1.md when interpreting validity windows. Search rank and graph centrality are navigation signals. Candidate contradictions need judgment; a structural gap does not establish a scientific research gap. Supersede an obsolete relation instead of erasing its history.
Explore: generate the self-contained viewer or export the graph as documented in references/cli.v1.md. Numbered source passages support an answer; community summaries are navigation aids. Do not send private graph content to a provider without authorization. A host completion command is optional; deterministic graph reads require no API key.
The graph library, analysis algorithms, SQLite store, adapters, viewer and existing records are retained. Knowledge owns graph relationships; it does not absorb the separate general storage capability or silently migrate Memory.
Limits and prerequisites
- Cooccurrence, centrality and candidate contradictions are not proof that a claim is true.
Your first useful run.
Start with a small source collection and one relationship you want to understand.
The download contains the complete skill folder. Extract it into the skills directory supported by your agent, keeping its references and license together. Or use the source installer below. Python 3.11+ is required for that installer.
Start from the public source
git clone https://github.com/michaeljabbour/amplifier-smart-tool-knowledge.git
cd amplifier-smart-tool-knowledgepython3 scripts/capability.py install --dest ~/.agents/skillsNative tools and optional execution
python3 -m pip install .These commands are for the optional native surface. Follow its installation guide and prerequisites; portable method installation is separate.
Keep the complete skill folder together, including its references and license. The host applies the method in your context; the skill does not start a background service.
- Python 3.11 or later is required by the portable install/build helper. The installed method runs through your agent host.
- The deterministic library and CLI use Python and SQLite without a model key.
- Native package and command identity stay knowledge; optional model-backed operations need a separately configured route.
- Choose KNOWLEDGE_HOME explicitly when you want a disposable test graph.
Try this with your agent
Use Knowledge to show how two fictional project notes support or challenge the same claim. Start with deterministic ingestion or structured additions. Trace each relationship to its source and distinguish an asserted fact from a candidate contradiction. Do not configure an external provider.Use after installing in a host that loads skill folders. Supply your own task files and review the result.
Keep the guide nearby
Read the guide · License · Changes · Release metadata · CONSOLIDATION
Know what you can rely on.
Native graph, storage, analysis and interface tests passed. A passing graph operation does not establish that an extracted claim or model-written answer is correct.
Read the checks and their scope
- Functional acceptance — passed
80 native tests passed after canonical distribution and test-package isolation updates.
- Native host acceptance — not-run
No live model, real desktop transaction, invitation, deployment or production data migration was performed by this functional baseline.
Source and conformance checks are separate from demonstrated benefit. Comparative effectiveness and outside-user outcomes are not established by this site.
Fictional examples explain the capability. Tests check the stated behavior. Outside-user outcomes need their own evidence.