OpenAI: reduce repeated heartbeat alerts by pashpashpash · Pull Request #65148 · openclaw/openclaw

pashpashpash · 2026-04-12T03:38:26Z

Summary

Problem: the OpenAI heartbeat overlay strongly nudged GPT-5 toward repeated user-facing heartbeat notifications.
Why it matters: in practice that can turn unchanged heartbeat state into low-value repeated alerts instead of more tasteful proactive behavior.
What changed: softened the heartbeat overlay by removing the explicit notify/stay-quiet instructions and replacing them with a narrower anti-repeat warning.
What did NOT change (scope boundary): this does not change the core heartbeat contract, delivery routing, or dedupe logic outside the OpenAI overlay.

Change Type (select all)

Scope (select all touched areas)

Linked Issue/PR

Closes #
Related #
This PR fixes a bug or regression

Root Cause (if applicable)

Root cause: the OpenAI overlay explicitly told heartbeat turns to be opinionated and notify when something seemed important, which could reinforce repeated unchanged alerts.
Missing detection / guardrail: the overlay did not discourage duplicate unchanged notifications.
Contributing context (if known): heartbeats already have a separate core reply contract, so the overlay-level notify bias was stronger than necessary.

Regression Test Plan (if applicable)

Coverage level that should have caught this:
- Unit test
- Seam / integration test
- End-to-end test
- Existing coverage already sufficient
Target test or file: extensions/openai/index.test.ts
Scenario the test should lock in: the friendly OpenAI overlay includes the revised heartbeat guidance and no longer asserts the old explicit notify/stay-quiet wording.
Why this is the smallest reliable guardrail: the change is pure prompt text in the OpenAI overlay.
Existing test that already covers this (if any): extensions/openai/index.test.ts
If no new test is added, why not: existing focused coverage was updated.

User-visible / Behavior Changes

GPT-5/OpenAI heartbeat guidance is less likely to encourage repeated unchanged alerts.

Diagram (if applicable)

N/A

Security Impact (required)

New permissions/capabilities? (Yes/No) No
Secrets/tokens handling changed? (Yes/No) No
New/changed network calls? (Yes/No) No
Command/tool execution surface changed? (Yes/No) No
Data access scope changed? (Yes/No) No
If any Yes, explain risk + mitigation:

Repro + Verification

Environment

OS: macOS
Runtime/container: local repo checkout
Model/provider: openai-codex/gpt-5.4
Integration/channel (if any): N/A
Relevant config (redacted): N/A

Steps

Inspect the OpenAI heartbeat overlay text.
Update the heartbeat wording to reduce explicit alert bias and add anti-repeat guidance.
Run the focused overlay test.

Expected

The overlay reflects the new heartbeat wording.
The focused OpenAI overlay test passes.

Actual

Matches expected.

Evidence

Attach at least one:

Failing test/log before + passing after
Trace/log snippets
Screenshot/recording
Perf numbers (if relevant)

Human Verification (required)

Verified scenarios: inspected the final overlay text and ran pnpm test extensions/openai/index.test.ts after rebasing onto latest origin/main.
Edge cases checked: verified the focused test assertion matches the revised heartbeat copy.
What you did not verify: end-to-end heartbeat behavior in a live session.

Review Conversations

I replied to or resolved every bot review conversation I addressed in this PR.
I left unresolved only the conversations that still need reviewer or maintainer judgment.

Compatibility / Migration

Backward compatible? (Yes/No) Yes
Config/env changes? (Yes/No) No
Migration needed? (Yes/No) No
If yes, exact upgrade steps:

Risks and Mitigations

Risk: prompt-only changes can shift behavior in ways the focused test cannot fully capture.
- Mitigation: scope is limited to the OpenAI overlay heartbeat block, and the core heartbeat contract is unchanged.

greptile-apps · 2026-04-12T03:40:10Z

Greptile Summary

This PR softens the heartbeat guidance in the OPENAI_FRIENDLY_PROMPT_OVERLAY constant by removing explicit notify/stay-quiet instructions and adding a concise anti-repeat warning ("Avoid spamming the user with the same notification over and over again."). The corresponding test in extensions/openai/index.test.ts is updated to assert the revised wording. No logic, routing, or dedupe changes are involved — this is a pure prompt-text edit scoped entirely to the OpenAI extension overlay.

Confidence Score: 5/5

Safe to merge — minimal, well-scoped prompt-text change with matching test coverage.

The change is entirely contained to the heartbeat section of the OpenAI overlay constant and its test; no logic, routing, or contract surfaces are touched. All test assertions in both the updated and pre-existing test cases align with the current overlay text, and no P0/P1 issues were found.

No files require special attention.

_{Reviews (1): Last reviewed commit: "OpenAI: reduce repeated heartbeat alerts" | Re-trigger Greptile}

aisle-research-bot · 2026-04-12T06:14:54Z

🔒 Aisle Security Analysis

We found 1 potential security issue(s) in this PR:

#	Severity	Title
1	🟠 High	Symlink-based file disclosure via unguarded HEARTBEAT.md read in heartbeat runner

1. 🟠 Symlink-based file disclosure via unguarded HEARTBEAT.md read in heartbeat runner

Property	Value
Severity	High
CWE	CWE-59
Location	`src/infra/heartbeat-runner.ts:560-566`

Description

The heartbeat runner reads HEARTBEAT.md directly from the agent workspace using fs.readFile() without the boundary/symlink protections used elsewhere for workspace bootstrap files.

Input/control: An attacker (or untrusted integration) able to write to the workspace directory can replace HEARTBEAT.md with a symlink (or other link) to an arbitrary file on disk (e.g., /etc/passwd, SSH keys, cloud credentials).
Sink: heartbeat-runner.ts reads the file content and may embed parts of it into the LLM prompt (Additional context from HEARTBEAT.md), which can then be exfiltrated via model output or used to influence tool behavior.
Missing defense: Unlike loadWorkspaceBootstrapFiles() which uses openBoundaryFile()/openVerifiedFileSync() (rejects symlinks/hardlinks, enforces root boundary and size limits), the heartbeat preflight bypasses these guards.

Vulnerable code:

const heartbeatFilePath = path.join(workspaceDir, DEFAULT_HEARTBEAT_FILENAME);
heartbeatFileContent = await fs.readFile(heartbeatFilePath, "utf-8");

Recommendation

Use the existing boundary-safe file open/read helpers (the same pattern as loadWorkspaceBootstrapFiles()), which validate the path is within workspaceDir and reject symlinks/hardlinks.

For example:

import syncFs from "node:fs";
import { openBoundaryFile } from "../infra/boundary-file-read.js";

const opened = await openBoundaryFile({
  absolutePath: heartbeatFilePath,
  rootPath: workspaceDir,
  boundaryLabel: "workspace root",
  maxBytes: 2 * 1024 * 1024,
});
if (opened.ok) {
  try {
    heartbeatFileContent = syncFs.readFileSync(opened.fd, "utf-8");
  } finally {
    syncFs.closeSync(opened.fd);
  }
}

Alternatively, refactor to call loadWorkspaceBootstrapFiles(workspaceDir) (or a shared helper) so all workspace reads go through the same guardrails.

Analyzed PR: #65148 at commit 728ced8

_{Last updated on: 2026-04-12T21:21:37Z}

openclaw-barnacle Bot added extensions: openai size: XS maintainer Maintainer-authored PR labels Apr 12, 2026

pashpashpash force-pushed the codex/openai-heartbeat-repeat-alerts branch from 96884e3 to 4b3b3fd Compare April 12, 2026 06:02

openclaw-barnacle Bot added extensions: memory-core Extension: memory-core size: S and removed size: XS labels Apr 12, 2026

This comment has been minimized.

Sign in to view

openclaw-barnacle Bot added channel: msteams Channel integration: msteams gateway Gateway runtime scripts Repository scripts labels Apr 12, 2026

openclaw-barnacle Bot added docs Improvements or additions to documentation size: M and removed size: S labels Apr 12, 2026

pashpashpash force-pushed the codex/openai-heartbeat-repeat-alerts branch from a6d236b to fc1b56a Compare April 12, 2026 06:48

openclaw-barnacle Bot added size: S and removed docs Improvements or additions to documentation channel: msteams Channel integration: msteams gateway Gateway runtime extensions: memory-core Extension: memory-core scripts Repository scripts size: M labels Apr 12, 2026

pashpashpash force-pushed the codex/openai-heartbeat-repeat-alerts branch from fc1b56a to 321176f Compare April 12, 2026 08:30

OpenAI: strengthen heartbeat overlay guidance

728ced8

pashpashpash force-pushed the codex/openai-heartbeat-repeat-alerts branch from 321176f to 728ced8 Compare April 12, 2026 21:19

pashpashpash merged commit f5447aa into main Apr 12, 2026
42 checks passed

pashpashpash deleted the codex/openai-heartbeat-repeat-alerts branch April 12, 2026 21:47

github-actions Bot mentioned this pull request Apr 12, 2026

📡 Upstream Digest — 2026-04-12 22:25 UTC curtismercier/openclaw-mods#554

Open

100yenadmin mentioned this pull request Apr 22, 2026

[Plan Mode 9/9] (OPTIONAL) GPT-5 prompt foundation #69449

Closed

lovewanwan pushed a commit to lovewanwan/openclaw that referenced this pull request Apr 28, 2026

OpenAI: strengthen heartbeat overlay guidance (openclaw#65148)

1ee31d0

ogt-redknie pushed a commit to ogt-redknie/OPENX that referenced this pull request May 2, 2026

OpenAI: strengthen heartbeat overlay guidance (openclaw#65148)

85c730e

github-actions Bot pushed a commit to Desicool/openclaw that referenced this pull request May 9, 2026

OpenAI: strengthen heartbeat overlay guidance (openclaw#65148)

6a225a4

globalcaos pushed a commit to globalcaos/tinkerclaw that referenced this pull request May 13, 2026

OpenAI: strengthen heartbeat overlay guidance (openclaw#65148)

61017b3

github-actions Bot pushed a commit to Desicool/openclaw that referenced this pull request May 24, 2026

OpenAI: strengthen heartbeat overlay guidance (openclaw#65148)

a20634b

jameslcowan pushed a commit to jameslcowan/openclaw that referenced this pull request Jun 2, 2026

OpenAI: strengthen heartbeat overlay guidance (openclaw#65148)

9c96341

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Uh oh!

OpenAI: reduce repeated heartbeat alerts#65148

OpenAI: reduce repeated heartbeat alerts#65148
pashpashpash merged 1 commit into
mainfrom
codex/openai-heartbeat-repeat-alerts

pashpashpash commented Apr 12, 2026

Uh oh!

greptile-apps Bot commented Apr 12, 2026

Uh oh!

This comment has been minimized.

aisle-research-bot Bot commented Apr 12, 2026 •

edited

Loading

Description

Recommendation

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Uh oh!

Uh oh!

Conversation

pashpashpash commented Apr 12, 2026

Summary

Change Type (select all)

Scope (select all touched areas)

Linked Issue/PR

Root Cause (if applicable)

Regression Test Plan (if applicable)

User-visible / Behavior Changes

Diagram (if applicable)

Security Impact (required)

Repro + Verification

Environment

Steps

Expected

Actual

Evidence

Human Verification (required)

Review Conversations

Compatibility / Migration

Risks and Mitigations

Uh oh!

greptile-apps Bot commented Apr 12, 2026

Greptile Summary

Confidence Score: 5/5

Uh oh!

This comment has been minimized.

aisle-research-bot Bot commented Apr 12, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

🔒 Aisle Security Analysis

Description

Recommendation

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

aisle-research-bot Bot commented Apr 12, 2026 •

edited

Loading