Skip to main content
Home/Tools/Assumption Ledger

Assumption Ledger

Assumption Ledger

A minimal, manifesto-aligned tool for agents to declare, revisit, and resolve assumptions. Provides traceable skepticism without auto-inference, scoring, or retry logic. Integrates with SequentialThinking and Workflow sessions via context references.

When to Use This Tool

  • Before SequentialThinking: Log dominant assumptions alongside initial prompt to establish context
  • During Workflow Execution: Reference ledger entries when decisions depend on unvalidated assumptions
  • After Completion: Update status to show what held true, capturing lessons in memory
  • Multi-agent Coordination: Share assumptions across agent boundaries for continuity
  • Risk Management: Make implicit assumptions explicit and trackable

Key features:

  • Captures assumption statements without inference or analysis
  • Stores as durable Markdown artifacts in memory://assumptions/
  • Integrates with knowledge graph through [[concept]] tags
  • Searchable via SearchMemories after Sync()
  • Supports status tracking (active, validated, invalidated, refined)
  • Maintains complete audit trail with timestamps
  • Zero auto-reasoning - agents decide what to validate and when

What This Tool Does NOT Do

In alignment with maenifold philosophy:

  • ❌ Does NOT automatically discover or rank assumptions
  • ❌ Does NOT cache, score, or batch operations beyond markdown writes
  • ❌ Does NOT implement telemetry, analytics, or usage tracking
  • ❌ Does NOT provide helper UI, wizards, or configuration generators
  • ❌ Does NOT infer relationships or create automated ontologies

Why: Every decision we make removes a decision the agent could make. The ledger is a mirror, not a mind.

Parameters

action (required)

One of: "append", "update", or "read"

For action="append" (Create New Assumption)

Required:

  • assumption: The assumption statement (free text)
  • concepts: Array of concept tags for knowledge graph integration (e.g., ["workflow", "sequential-thinking"])

Optional:

  • context: Reference to workflow or sequential thinking session (e.g., "workflow://thinking/session-1756610546730")
  • validationPlan: How you plan to validate this assumption (free text)
  • confidence: Free text confidence level (e.g., "high", "medium", "low", "needs-verification")

For action="update" (Modify Existing Assumption)

Required:

  • uri: Memory URI of the assumption (e.g., "memory://assumptions/2025/09/assumption-1759186965105")

Optional:

  • status: New status - one of: "active", "validated", "invalidated", or "refined"
  • confidence: Updated confidence level (free text)
  • validationPlan: Updated validation plan
  • notes: Additional notes to append (creates timestamped section)

For action="read" (View Existing Assumption)

Required:

  • uri: Memory URI of the assumption

Output Format

Append

Returns confirmation with:

  • Memory URI for the new assumption
  • Statement, confidence, and status summary
  • Declarative next-step suggestion (run Sync() for search integration)
  • Validation plan reminder if provided

Update

Returns confirmation with:

  • Updated URI
  • New status and/or confidence if changed
  • Declarative next-step suggestion (run Sync() to update graph)

Read

Returns full assumption details:

  • All frontmatter metadata (status, confidence, timestamps, context, validation plan)
  • Complete content including statement, validation plan, and concept links
  • Update history if notes have been added

Integration Patterns

With SequentialThinking

Before starting session:
AssumptionLedger(
  action: "append",
  assumption: "Current implementation uses recursive traversal",
  context: "session-1759186950",
  concepts: ["graph-traversal", "performance"],
  validationPlan: "Profile actual query performance"
)

Then in thoughts:
"Analyzing [[graph-traversal]] performance assuming [[recursive-implementation]]...
See memory://assumptions/2025/09/assumption-X for baseline assumption"

With Workflow

AssumptionLedger(
  action: "append",
  assumption: "Bug fix requires schema migration",
  context: "workflow://thinking/workflow-1759186950",
  concepts: ["database", "migration"]
)

With SearchMemories

# First, sync assumptions to knowledge graph
Sync()

# Then search across all assumptions
SearchMemories(
  query: "workflow orchestration",
  mode: "Hybrid",
  folder: "assumptions"
)

File Structure

Assumptions stored at: memory://assumptions/YYYY/MM/assumption-{timestamp}.md

Each file contains:

  1. YAML Frontmatter: status, confidence, context, validation_plan, timestamps
  2. Markdown Content: Human-readable assumption statement and related concepts

Example:

---
created: 2025-09-29T23:02:45.1054396Z
status: active
confidence: medium
context: workflow://thinking/session-1756610546730
validation_plan: Validate once dialogue MCP hooks are reintroduced
---

# Assumption: The dialogue tool will remain MCP-only

## Statement

The dialogue tool will remain MCP-only

## Validation Plan

Validate once dialogue MCP hooks are reintroduced

## Related Concepts

- [[dialogue]]
- [[workflow-dispatch]]

Workflow Guidance

Before SequentialThinking

Log dominant assumptions alongside the initial prompt to establish context.

During Workflow Execution

Reference ledger entries in thoughts when a decision depends on them.

After Completion

Update status to show what held true, capturing lessons in memory:// rather than proposing new features.

Notes

  • Requires [[concept]] tags for knowledge graph integration
  • Files persist across sessions in date-organized folders
  • Searchable after running Sync()
  • Updates create timestamped sections maintaining complete history
  • Zero automation beyond file write - validation is agent's responsibility

See Also

← Back to Tools↑ Back to Top