Proposed specificationSource-located exemplarH3

Propose-Probe-Compose

Improve a harness using contrastive probes.

JOIN POINTharness_edit_cycleRECORD STATUSCandidate · v0.1.0

UML class view

The proposed contract drawn as participant roles, received calls, and numbered dependencies.

Object interactions

The same roles and calls as a class collaboration and a numbered message sequence.

Object interaction diagram for Propose-Probe-Compose, showing role collaboration and a numbered message sequence — FIG. E02 / OBJECT INTERACTIONS
Open full SVG Download PDF

Pointcut

Where the mechanism applies. All four dimensions belong to the proposed match, not just its vocabulary.

Context: A base harness is available for improvement.
Trigger: A proposal changes a bounded harness component and probes are executed.
Structure: Proposal scope, task probes, evaluator, and composition operation.
Effect: Only evidence-supported edits are composed into the next harness.

Protected contract

The obligation that clients, operators, and auditors should be able to rely on.

Requires: Harness components are addressable and probes use permitted data.
Ensures: Candidate edits carry probe results and composition provenance.
Invariant: Evaluation data cannot be silently repurposed as final test data.
Violation: Failed probes, invalid edits, and conflicting changes remain uncomposed.

Negative control

A one-shot prompt rewrite without probes is not E02.

Variation & lifecycle

The hot spot is admitted only where the protected semantics survive.

Permitted variation

Contrastive probes, recursive improvement, and component-level editing.

Test obligation

Test accepted edit, rejected edit, leakage, and composition rollback.

Evolution question

Did harness edits become evidence-gated and attributable?

Prompt + code design

An illustrative prompt and a small design sketch for the proposed contract. Names and APIs are invented; this is not a repository excerpt.

PROMPT SKETCH

Propose a harness edit, probe it on permitted cases, and compose only edits whose probe results support the change. Keep the proposed diff, probe evidence, and provenance together.

CODE DESIGN / PYTHON

def improve(harness, probes):
    proposal = proposer.edit(harness)
    results = [probe.run(proposal.candidate) for probe in probes]
    record = EditRecord(proposal.diff, results, harness.version)
    if not all(result.accepted for result in results):
        return Rejected(record)
    return Composed(harness.apply(proposal), record)

Illustrative design sketch. A probe is evidence about this edit, not a universal guarantee for the next one. This example is a teaching aid, not implementation evidence or a conformance claim.

Detection & matching

The frozen rule that retrieves this record, the canonical source exemplar, and the control that should stay unmatched.

Join point: harness_edit_cycle
Source locator: Xu et al.
Lexical cue: (?i)\bpropose[_-]?probe[_-]?compose\b|\bproposal\b.*\bprobe\b|\bprobe\b.*\bcompose\bFrozen expression, v0.1.0. A match is a retrieval cue, not a confirmed implementation.
Aliases: proposeprobecompose, candidateprobe, harnesseditcycle
Canonical exemplar: google-research/harness2harness2/core/recursion.py:360Support: tests/test_workflow.py:139
Negative control: A one-shot prompt rewrite without probes is not E02.

Scope-anchored review. pohe-agent analyze <repo> --pattern E02 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 ↗

Repository files

Up to five public files retrieved by the frozen cue and pinned to the audited commit. A match is a place to inspect, not a confirmed implementation.

01google-research/harness2harness2/core/recursion.py:360CANONICAL SOURCE 02langflow-ai/langflowsrc/lfx/src/lfx/services/settings/groups/runtime.py:216RETRIEVAL CUE 03QwenLM/qwen-codepackages/cli/src/serve/routes/health.ts:32RETRIEVAL CUE 04affaan-m/ECCdocs/design/context-profile-ai-evaluation.md:94RETRIEVAL CUE 05bytedance/deer-flowbackend/tests/support/compose.py:24RETRIEVAL CUE

Evidence & provenance

Each layer answers a different question. Do not substitute retrieval for review.

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 COVERAGE6/36

16.7% 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: 6 / 41

Canonical source anchorgoogle-research/harness2 686881751887harness2/core/recursion.py:360

Supporting surfacetests/test_workflow.py:139

Proposal, probe/evaluation, and composition rounds.

Exploratory retrieval cue Regex / v0.1.0

(?i)\bpropose[_-]?probe[_-]?compose\b|\bproposal\b.*\bprobe\b|\bprobe\b.*\bcompose\b

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