Proposed specificationSource-located exemplarH2

Verify-on-stop guards

Check artifacts before accepting completion.

JOIN POINTterminal_acceptance_interceptionRECORD 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 Verify-on-stop guards, showing role collaboration and a numbered message sequence — FIG. P17 / 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: The agent proposes completion after producing an artifact.
Trigger: A stop or finish action is requested.
Structure: Exit interception, verifier execution, and accept-or-continue result.
Effect: The stop is accepted only after the verifier reports sufficient evidence.

Protected contract

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

Requires: Stop can be intercepted and at least one verification is available.
Ensures: Failed verification returns actionable feedback or blocks completion.
Invariant: The same acceptance gate cannot be bypassed by another stop path.
Violation: Unavailable or flaky verifier yields explicit unknown or failure, not silent success.

Negative control

Tests run by the user after the run are not verify-on-stop.

Variation & lifecycle

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

Permitted variation

Project test commands, acceptance hooks, verifier agents, and blocking policies.

Test obligation

Test pass, fail-and-continue, verifier error, repeated stop, and bypass route.

Evolution question

Did completion verification become mandatory rather than advisory?

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

Do not accept completion until the artifact has passed the configured verification. A failed check must return actionable feedback or block the stop; an unavailable verifier is unknown, never success.

CODE DESIGN / PYTHON

def guarded_stop(agent, proposal, verifier):
    if not proposal.requests_stop:
        return agent.continue_run()
    verdict = verifier.check(proposal.artifact)
    if verdict.passed:
        return agent.accept(proposal, evidence=verdict.evidence)
    if verdict.unknown:
        return agent.block(proposal, reason="verification unavailable")
    return agent.continue_run(feedback=verdict.feedback)

Illustrative design sketch. The gate, not the completed status, is the acceptance authority. 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: terminal_acceptance_interception
Source locator: Barbaste et al., §6.2, p. 15 (Hermes stop-guards); Table 12 “Outer Verification Loop”, p. 53; source label
Lexical cue: (?i)\b(?:verify[_-]?on[_-]?stop|before[_-]?stop|stop[_-]?verif|completion[_-]?check|acceptance[_-]?check)\bFrozen expression, v0.1.0. A match is a retrieval cue, not a confirmed implementation.
Aliases: verifyonstop, beforestop, stopverify, completioncheck, acceptancecheck
Canonical exemplar: Darwin-Agent/HarnessXharnessx/processors/control/self_verify.py:95Support: tests/unit/test_processor_triggers.py:336
Negative control: Tests run by the user after the run are not verify-on-stop.

Scope-anchored review. pohe-agent analyze <repo> --pattern P17 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.

01Darwin-Agent/HarnessXharnessx/processors/control/self_verify.py:95CANONICAL SOURCE 02agno-agi/agnolibs/agno/agno/os/routers/schedules/router.py:61RETRIEVAL CUE 03mastra-ai/mastrapackages/core/src/stream/base/output.ts:253RETRIEVAL CUE 04openclaw/openclawsrc/agents/embedded-agent-runner/run/attempt-stream-prepare.ts:105RETRIEVAL CUE 05stablyai/orcasrc/main/native-chat/agent-session-wire/recovery/structured-agent-session-restart-resume-host.ts:294RETRIEVAL 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 COVERAGE8/36

22.2% 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: 7 / 41

Canonical source anchorDarwin-Agent/HarnessX ccb411691343harnessx/processors/control/self_verify.py:95

Supporting surfacetests/unit/test_processor_triggers.py:336

Task-end self-verification processor.

Exploratory retrieval cue Regex / v0.1.0

(?i)\b(?:verify[_-]?on[_-]?stop|before[_-]?stop|stop[_-]?verif|completion[_-]?check|acceptance[_-]?check)\b

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