hypothesis

A command for recording ideas that can be tested as hypotheses in a connected knowledge graph. Each hypothesis moves from proposed to testing, then validated or rejected as evidence is added.

In plain words
What is it for?
Use it to create hypotheses, find related ones, record validation plans, and link evidence to claims across projects.
Why use it?
It keeps testable claims, expected results, and supporting evidence together instead of leaving them scattered across notes. This makes it easier to check what is known and what still needs testing.

Command for Claude Code

Part of the cortex plugin — 1 skill, 15 commands, 3 agents, 5 hooks, 2 MCP servers shipped together

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add commands/peaky8linders/claude-cortex/hypothesis
Clone the repo
git clone --depth 1 https://github.com/Peaky8linders/claude-cortex

Made for: Claude Code.

Or install cortex, the plugin that ships this one along with the rest of its 1 skill, 15 commands, 3 agents, 5 hooks, 2 MCP servers.

Per session 13 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 758 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00013 $0.00758
Opus 5 $0.00006 $0.00379
Sonnet 5 $0.00003 $0.00152
Haiku 4.5 $0.00001 $0.00076

Measured 3d ago against content hash b11b9389b3db, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

hypothesis scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 3d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

.claude/commands/hypothesis.md · 94 lines

How it starts

The opening of the file, as written. The whole thing — 94 lines — stays where its author put it; the contents beside it link to each section on GitHub.

/hypothesis — Manage Learning Hypotheses via Knowledge Graph

You manage hypotheses in the Brainiac knowledge graph. Hypotheses are graph nodes that move through: proposed -> testing -> validated/rejected. Evidence is tracked as causal edges from evidence nodes to the hypothesis node.

System Location

  • Engine: ~/.claude/knowledge/brainiac/
  • CLI: cd ~/.claude/knowledge && python -m brainiac <command>

Commands

Parse the user's intent from their message after /hypothesis:

Create a new hypothesis

  1. Search existing graph for related hypotheses:

    cd ~/.claude/knowledge && python -m brainiac search "TOPIC"
    
  2. Add hypothesis as a graph node:

    cd ~/.claude/knowledge && python -m brainiac add hypothesis "CLAIM: [testable statement]. EXPECTED: [what we'd observe]. TEST: [how to validate]."
    
  3. Also create ~/.claude/knowledge/hypotheses/{kebab-name}.md with full detail:

    ---
    name: descriptive-name
    type: hypothesis
    projects: [relevant-projects]
    tags: [relevant-tags]
    created: YYYY-MM-DD
    updated: YYYY-MM-DD
    confidence: low
    status: proposed
    graph_id: <node ID from step 2>
    ---
    
    ## Claim
    [Clear, testable statement]
    
    ## Expected Outcome
    [What would we observe if this is true?]
    
    ## How to Test
    [Concrete steps to validate or reject]
    
    ## Evidence For
    <!-- Add dated entries as evidence accumulates -->
    
    ## Evidence Against
    <!-- Add dated entries as evidence accumulates -->
    
    ## Verdict
    Pending — needs more evidence.
    
  4. Update ~/.claude/knowledge/hypotheses/INDEX.md with the new entry.

Add evidence to an existing hypothesis

  1. Search for the hypothesis: cd ~/.claude/knowledge && python -m brainiac search "hypothesis topic"
  2. Update the hypothesis markdown file with dated evidence entry
  3. Create a causal edge from the evidence source to the hypothesis:
    cd ~/.claude/knowledge && python -m brainiac link <evidence_node_id> <hypothesis_node_id> causal
    
  4. Update confidence (low/medium/high) based on accumulated evidence
  5. If 3+ data points consistently one way:
    • Validated: Set status to validated. Create a pattern/decision node via /learn. Link hypothesis -> pattern with causal edge.
    • Rejected: Set status to rejected. Create an antipattern node via /learn. Link hypothesis -> antipattern with causal edge.

Read the full file on GitHub · 94 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 3d ago First seen · 94 lines · 13 tokens per session scan A b11b9389b3db

Subscribe to this mod's changes

hypothesis is a command published in the GitHub repository Peaky8linders/claude-cortex (11 stars, last pushed 2mo ago), licensed MIT. It adds 13 tokens to every session and 758 once invoked, about $0.0001 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.