What happens when a hypothesis expires?

Expiration is based on claim type (e.g., financial 90 days, technology 180 days, strategy 365 days, leadership no expiration). If expired, query with include_expired=True and then re-validate, invalidate, or extend.

How do I add a new hypothesis?

Use save_hypothesis with a unique hypothesis_id, include a clear claim, topic, initial confidence (untested), and attach evidence if available.

How do I filter results by confidence or topic?

Use get_hypotheses with parameters like company, topic, confidence, and include_expired as needed to refine results.

hypothesis-tracking

Scanned

npx machina-cli add skill blisspixel/primr/hypothesis-tracking --openclaw

Files (1)

SKILL.md

3.4 KB

Hypothesis Tracking

Primr uses a four-level confidence system: UNTESTED, VALIDATED, INVALIDATED, CONFIRMED. See references/confidence-levels.md for the full framework and lifecycle.

Operational Capabilities

Retrieve Hypotheses

get_hypotheses(company="Acme Corp")
get_hypotheses(company="Acme Corp", confidence="validated")
get_hypotheses(company="Acme Corp", topic="technology")
get_hypotheses(company="Acme Corp", include_expired=True)

Update Hypothesis Confidence

save_hypothesis(company="Acme Corp", hypothesis_id="h_001",
    confidence="validated", evidence="Found AWS case study on company blog")

save_hypothesis(company="Acme Corp", hypothesis_id="h_002",
    confidence="invalidated", evidence="CEO states they use Azure exclusively")

save_hypothesis(company="Acme Corp", hypothesis_id="h_001",
    confidence="confirmed", evidence="Multiple sources: job posts, blog, press release")

Create New Hypothesis

save_hypothesis(company="Acme Corp", hypothesis_id="h_new_001",
    claim="Company is expanding into healthcare",
    confidence="untested", topic="strategy")

Expiration Rules

Financial claims: 90 days
Technology claims: 180 days
Strategy claims: 365 days
Leadership claims: no expiration

When expired hypotheses exist, query with include_expired=True, then re-validate, invalidate, or extend.

Error Handling

Error	Resolution
Hypothesis not found	Verify company name and hypothesis_id
Duplicate ID	Use a unique hypothesis_id per company
Missing evidence	Always provide evidence string when changing confidence
Expired hypothesis	Re-query with `include_expired=True` to access

Example Workflow

User: "What do we know about Acme Corp's technology?"

1. get_hypotheses("Acme Corp", topic="technology")

   Results:
   - [CONFIRMED] Uses microservices architecture
     Evidence: CTO interview, job posts, tech blog
   - [VALIDATED] AWS as primary cloud
     Evidence: Job posting mentions AWS
   - [UNTESTED] Considering Kubernetes migration
     Evidence: None yet

2. Present to user:
   "Here's what we know about Acme Corp's technology:

   CONFIRMED: Uses microservices architecture
     Multiple sources confirm this

   VALIDATED: AWS as primary cloud
     One supporting source (job posting)

   UNTESTED: Considering Kubernetes migration
     No evidence yet, extracted from blog speculation

   Would you like me to research any of these further?"

3. If user provides new evidence:
   save_hypothesis(company="Acme Corp", hypothesis_id="h_003",
       confidence="validated",
       evidence="User confirmed: spoke with Acme engineer at conference")

Constraints

Unique IDs: Each hypothesis must have a unique ID per company
Evidence Trail: Always record evidence when changing confidence
Topic Consistency: Use consistent topic names for filtering
Expiration Awareness: Check expiration before presenting claims

Source

git clone https://github.com/blisspixel/primr/blob/main/skills/hypothesis-tracking/SKILL.mdView on GitHub

Overview

Hypothesis Tracking helps you store, filter, and track research hypotheses about a company using a four-level confidence system (UNTESTED, VALIDATED, INVALIDATED, CONFIRMED). It provides an evidence trail, expiration rules by claim type, and a workflow to re-validate or extend hypotheses as new data emerges.

How This Skill Works

Hypotheses are recorded per company with fields like hypothesis_id, claim, topic, confidence, and evidence. Use get_hypotheses to retrieve and filter, then save_hypothesis to update confidence with supporting evidence. The system enforces expiration based on claim type and maintains an audit trail of changes.

When to Use It

When a user asks about prior research on a company
When validating a specific claim (e.g., technology, leadership, strategy)
When tracking the lifecycle of claims and expiring ones
When compiling an evidence-backed summary for stakeholders
When introducing a new hypothesis and attaching initial evidence

Quick Start

Step 1: Retrieve or create hypotheses with get_hypotheses(company="Acme Corp") or save_hypothesis(...) for a new one
Step 2: Update confidence with evidence using save_hypothesis(company="Acme Corp", hypothesis_id="h_001", confidence="validated", evidence="Source...")
Step 3: Before presenting results, check expiration rules and re-validate or extend as needed

Best Practices

Ensure a unique hypothesis_id per company
Always attach evidence when saving or updating confidence
Keep topic names consistent for reliable filtering
Check expiration rules before presenting claims
Periodically re-validate or extend hypotheses as new data arrives

Example Use Cases

Acme Corp: CONFIRMED - Uses microservices architecture; evidence from CTO interview, job posts, tech blog
Acme Corp: VALIDATED - AWS as primary cloud; evidence from job posting
Acme Corp: UNTESTED - Kubernetes migration; evidence: none yet
Acme Corp: INVALIDATED - Azure-exclusive claim; evidence: CEO interview
Acme Corp: NEW HYPOTHESIS h_new_001 - 'Expanding into healthcare'; topic: strategy; confidence: untested

Frequently Asked Questions

Add this skill to your agents