Catalog/R family/R02
Proposed specificationSource-located exemplarH2H3

Verbal reflection memory

Reuse feedback across attempts.

JOIN POINTfeedback_reuse_across_attemptsRECORD STATUSCandidate · v0.1.0
UML class view for Verbal reflection memory, showing participant roles, received calls, and analytical dependencies
FIG. R02 / UML CLASS VIEW
Object interaction diagram for Verbal reflection memory, showing role collaboration and a numbered message sequence
FIG. R02 / OBJECT INTERACTIONS
Context
An agent receives feedback that can inform later attempts.
Trigger
A failure or critique is verbalized and stored.
Structure
Reflection producer, retrieval key, memory scope, and update policy.
Effect
A later attempt retrieves attributable feedback and uses it in context.
Requires
Attempt and feedback identities are linked.
Ensures
Retrieved reflection retains source and remains updateable or revocable.
Invariant
Reflected explanations are not promoted to verified facts.
Violation
Contradictory or poisoned reflections are filtered or explicit.

Negative control

A generic conversation history is not a reflection-memory contract.

Permitted variation

Reflexion, episodic memory, observational memory, and project notes.

Test obligation

Test reuse, stale memory, contradiction, poisoning, and deletion.

Evolution question

Did reflection memory gain provenance and maintenance policies?

PROMPT SKETCH
Store verbal reflections with links to the attempt and feedback that produced them. Retrieve a reflection only with its source, and keep it updateable or revocable when later evidence changes.
CODE DESIGN / PYTHON
def reflect(attempt, feedback, memory):
    note = Reflection(
        text=evaluator.explain(attempt, feedback),
        attempt_id=attempt.id,
        feedback_id=feedback.id,
        scope=attempt.task.scope,
    )
    memory.put(note)
    return note

def retrieve(task, memory):
    return [note for note in memory.search(task) if note.scope == task.scope]

Illustrative design sketch. A reflection without provenance becomes an unfalsifiable instruction. This example is a teaching aid, not implementation evidence or a conformance claim.

Join point
feedback_reuse_across_attempts
Source locator
See the source record.
Lexical cue
(?i)\b(?:Reflexion|verbal[_-]?reflection|reflection[_-]?memory|episodic[_-]?memory)\bFrozen expression, v0.1.0. A match is a retrieval cue, not a confirmed implementation.
Aliases
reflectionmemory, episodicmemory, lessonslearned, reflexion
Canonical exemplar
mastra-ai/mastrapackages/memory/src/processors/observational-memory/reflector-agent.ts:50Support: packages/memory/src/processors/observational-memory/__tests__/observational-memory.test.ts:10564
Negative control
A generic conversation history is not a reflection-memory contract.

Scope-anchored review. pohe-agent analyze <repo> --pattern R02 anchors the cue to its enclosing declaration and checks a separate operational signature for all four pointcut dimensions. Missing roles and negative-control risks remain visible; a structural match is not E2/E3. Calibration report ↗

Source-located exemplar

A concrete source anchor is available for this proposed record. It is one exemplar, not evidence that all repositories—or every contract field—conform.

Review boundary: this candidate and its specification come from an AI-assisted study. Independent human validation remains open.

ANY-EVIDENCE FAMILY COVERAGE1/36

2.8% of repository families in the frozen panel have at least one current cue, historical cue, or canonical source layer. This is not confirmed prevalence.

Current lexical-cue repositories: 0 / 41
Exploratory retrieval cue Regex / v0.1.0
(?i)\b(?:Reflexion|verbal[_-]?reflection|reflection[_-]?memory|episodic[_-]?memory)\b

This expression is a discovery instrument only. A match is a retrieval cue—not a confirmed implementation, pointcut match, or contract proof.