feat(cli): add agent markers other detectors use, verified upstream by jrusso1020 · Pull Request #7425 · heygen-com/hyperframes

jrusso1020 · 2026-11-09T02:49:18Z

What

Markers from other detectors, verified upstream. I compared our rules with std-env, @vercel/detect-agent, am-i-vibing and agent-cli-detector, and kept only markers the agent itself sets on the shells it spawns:

Agent	Marker	Evidence
Auggie (`auggie`, new)	`AUGMENT_AGENT`	`@augmentcode/auggie` bundle: the shell tool is built with `env: { AUGMENT_AGENT: "1" }`
Amp	`AMP_CURRENT_THREAD_ID`	`@sourcegraph/amp` bundle: the Bash tool and toolbox subprocesses get it
Cursor	`CURSOR_CONVERSATION_ID`, `CURSOR_SANDBOX`	CLI and IDE agent-exec bundles; `CURSOR_SANDBOX` documented in cursor.com run-modes
Cline CLI	`CLINE_WRAPPER_PATH`	`cline/cline` `apps/cli/bin/cline`
Muse Code (`muse`, new)	`MUSE_RELEASE_INFO`	Meta's `muse-launcher.sh` exports it before exec
Warp Oz (`warp_oz`, new)	`OZ_RUN_ID`	`warpdotdev/warp` `crates/warp_cli/src/lib.rs`. Checked after the harnesses it wraps (Claude Code, Codex, Gemini).

Claude Code fix. The rule now keys on CLAUDE_CODE_CHILD_SESSION (or the inherited CLAUDE_CODE_ENTRYPOINT), not CLAUDECODE. Claude Code's docs say its IDE extensions also set CLAUDECODE in their integrated terminals, so a person typing in that terminal counted as an agent: isUnattended stopped prompting them, and auth login took the agent path.
Package size budget: @hyperframes/cli packed budget goes from 7982002 to 8062664, the measured size. main already sat at the edge of the 1% tolerance, and these rules put the packed CLI 842 bytes past it.
Cleanup: fixes the stale "rules above" wording from fix(cli): let an agent's own name beat editor and host markers #7424 review, plus two citations the comment gate flags.

Why

Follow-up to #7423 / #7424.

Left out on purpose: user settings and tokens such as COPILOT_MODEL, COPILOT_ALLOW_ALL, GOOSE_PROVIDER, OPENCODE_CLIENT, JUNIE_DATA, FACTORY_ENV. Also left out:

markers set on a whole workspace a person also types in (CURSOR_TRACE_ID, REPLIT_SESSION, Bolt's, KILO_RUN_ID);
hook-only markers (DEVIN_PROJECT_DIR, DROID_PROJECT_DIR);
anything only seen in a compiled binary with no readable context (KIRO_SESSION_ID, ANTIGRAVITY_TRAJECTORY_ID).

Related work

Refs #7423, #7424

How

New rules sit with the other agent rules; warp_oz goes just before the Agent SDK rule.
Tests that simulated a Claude Code shell with CLAUDECODE=1 now use CLAUDE_CODE_CHILD_SESSION=1.
A new test pins that CLAUDECODE alone (the IDE terminal) is neither claude_code nor agent-driven.

Test plan

Unit tests: each new marker, Oz losing to the harness it wraps (Codex), and the Claude IDE terminal case.
Ran the telemetry, attendedTerminal, desktopApp, auth login, agentMode, telemetryIdentity and usage tests: 1136/1136. tsc --noEmit, comment-citations and comment-ratchet also pass.
Manual testing performed
Comments follow CONTRIBUTING.md "Comments"

— Rames

🤖 Generated with Claude Code

Compared std-env, @vercel/detect-agent, am-i-vibing and agent-cli-detector against our rules and kept only markers the agent itself sets on the shells it spawns: Auggie AUGMENT_AGENT, Amp AMP_CURRENT_THREAD_ID, Cursor CURSOR_CONVERSATION_ID and CURSOR_SANDBOX, Cline CLI CLINE_WRAPPER_PATH, Muse MUSE_RELEASE_INFO, and Warp Oz OZ_RUN_ID (after the harnesses it wraps). Claude Code now keys on CLAUDE_CODE_CHILD_SESSION or CLAUDE_CODE_ENTRYPOINT. Its IDE extensions also set CLAUDECODE in the integrated terminal a person types into, so CLAUDECODE alone counted that person as an agent: no prompts, agent-mode login. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

github-actions · 2026-11-09T02:58:56Z

Edit accuracy: accurate 267 (base branch 267), smooth 113 of those

The gate passes.
Smoothness is reported in the artifact, not gated. A case fails only if it fails 2 of 3 runs.

Quarantined, measured but not gated (8)

seqplainkeys-none-center-r0-nested-z100 (fixed by studio: undo/redo on a keyframed plain element sometimes leaves the box 230-400px off (edit-accuracy seqplainkeys-none-*) #6989): pass, passes
seqplainkeys-none-center-r0-root-z100 (fixed by studio: undo/redo on a keyframed plain element sometimes leaves the box 230-400px off (edit-accuracy seqplainkeys-none-*) #6989): pass, passes
seqplainkeys-none-center-r30-root-z100 (fixed by studio: undo/redo on a keyframed plain element sometimes leaves the box 230-400px off (edit-accuracy seqplainkeys-none-*) #6989): not run
seqplainkeys-none-pct-r0-root-z100 (fixed by studio: undo/redo on a keyframed plain element sometimes leaves the box 230-400px off (edit-accuracy seqplainkeys-none-*) #6989): pass, passes
seqplainkeys-none-pct-r30-root-z100 (fixed by studio: undo/redo on a keyframed plain element sometimes leaves the box 230-400px off (edit-accuracy seqplainkeys-none-*) #6989): not run
seqplainkeys-none-px-r0-nested-z100 (fixed by studio: undo/redo on a keyframed plain element sometimes leaves the box 230-400px off (edit-accuracy seqplainkeys-none-*) #6989): pass, passes
seqplainkeys-none-px-r30-nested-z100 (fixed by studio: undo/redo on a keyframed plain element sometimes leaves the box 230-400px off (edit-accuracy seqplainkeys-none-*) #6989): not run
seqplainkeys-none-px-r30-root-z100 (fixed by studio: undo/redo on a keyframed plain element sometimes leaves the box 230-400px off (edit-accuracy seqplainkeys-none-*) #6989): not run

Unstable (2)

seqrotatenudge-none-pct-r0-root-z100: tracking 0.03, pressJump 0, drop 0, reload 3.03, render 0.04, renderKey -, undo false, teleport true / tracking 0.03, pressJump 0, drop 0, reload 0, render 0.08, renderKey -, undo true, teleport true / tracking 0.03, pressJump 0, drop 0, reload 0, render 0.08, renderKey -, undo true, teleport true
seqab-none-px-r0-root-z100: tracking 0.03, pressJump 0, drop 0, reload 0, render 0.03, renderKey -, undo true, teleport true / tracking 0.03, pressJump 0, drop 0, reload 0, render 0.03, renderKey -, undo true, teleport true / tracking 0.03, pressJump 0, drop 0, reload 0, render 0.03, renderKey -, undo true, teleport true

main already sat at the edge of the 1% tolerance; the new agent rules put the packed CLI 842 bytes past it (8062664 vs 7982002 + 1%). Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

heygengenesis

Verdict: APPROVE · 0 blocker · 0 important · 2 nits · head 5d2ee2e3

The Claude Code change holds up against the 2.1.319 binary and VS Code extension. The Bash tool sets CLAUDE_CODE_CHILD_SESSION along with CLAUDECODE. claude puts CLAUDE_CODE_ENTRYPOINT in its own env at startup, so every child inherits it. The extension's terminals set neither marker. Both nits are comment wording ↓

Scope, blast radius, CI

isUnattended: a person whose IDE terminal has only CLAUDECODE now gets prompts again. Every real Claude Code spawn still matches through CHILD_SESSION or ENTRYPOINT, including stdio MCP servers and versions before 2.1.172, so agents still get no prompts.
auth login: agentShellOnThisMachine only matters when stdout is not a TTY, so the CLAUDECODE change has no effect on a TTY user. The new markers send more agents without a TTY to the browser sign-in, which is the intended result.
desktopApp: someoneToTell short-circuits on tty, so a person at a terminal sees no change. Agents with the new markers now get the hint. Two other callers of detectAgentRuntime also change behaviour: feedback.ts (an IDE-terminal person gets the normal TTY prompt instead of the agent line) and the Studio __HF_CLI_AGENT_RUNTIME label. Both changes are correct.
I found no downstream enum for AgentRuntime that needs auggie, muse or warp_oz added. usage.ts detectHarness still keys on CLAUDECODE, which is fine for that command.
Spot checks: Cline apps/cli/bin/cline sets CLINE_WRAPPER_PATH in childEnv. Warp's local harness panes are hidden and launched off-screen, so OZ_RUN_ID does not leak into a terminal a person types in. OpenCode, which Oz also wraps, is matched before warp_oz.
Local runs: telemetry/*, attendedTerminal, auth login, desktopApp, telemetryIdentity and usage gave 616/616. cli.agentMode gave 518/518. I ran the suites both with and without Claude Code markers in the environment.
CI at 5d2ee2e3: CI, Windows render verification, regression, CodeQL and Comments are all green. The budget bump is explained in the PR body.

— Genesis

heygengenesis · 2026-11-09T03:27:29Z

-  // every child (main.tsx:527). Both propagate to spawned subprocesses.
-  // Source: confirmed by @magi from Claude Code internal source.
+  // Claude Code sets CLAUDE_CODE_CHILD_SESSION on what its tools spawn, and an
+  // inherited CLAUDE_CODE_ENTRYPOINT (older versions). Not CLAUDECODE: its IDE


nit: CLAUDE_CODE_ENTRYPOINT is not only from older versions. Current claude (2.1.319) sets it to cli/sdk-cli/mcp at startup. It is also the only Claude marker that reaches a stdio MCP server (Claude Code removes CHILD_SESSION there) and shells from versions before 2.1.172. Calling it "older versions" invites someone to delete it later, which would break detection in exactly those cases.

Suggested change

// inherited CLAUDE_CODE_ENTRYPOINT (older versions). Not CLAUDECODE: its IDE

// inherited CLAUDE_CODE_ENTRYPOINT (MCP servers, pre-2.1.172). Not CLAUDECODE: its IDE

heygengenesis · 2026-11-09T03:27:29Z

+  // Source: https://github.com/warpdotdev/warp (crates/warp_cli/src/lib.rs)
+  { name: "warp_oz", check: (env) => typeof env["OZ_RUN_ID"] === "string" },
  // An app built on the Claude Agent SDK, running a command outside Claude
  // Code's own shell tool (that path sets CLAUDECODE, caught above). The SDK


nit: this comment is out of date now that the rule no longer reads CLAUDECODE.

Suggested change

// Code's own shell tool (that path sets CLAUDECODE, caught above). The SDK

// Code's own shell tool (that path sets CLAUDE_CODE_CHILD_SESSION, caught above). The SDK

CLAUDE_CODE_ENTRYPOINT is set by current Claude Code versions too, and it is the only Claude marker MCP servers inherit, so it is not an "older versions" fallback. The Agent SDK comment still said Claude Code's shell tool sets CLAUDECODE; the rule keys on CLAUDE_CODE_CHILD_SESSION. Follow-up to review nits on #7425. Comments only. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

…ygen-com#7426) CLAUDE_CODE_ENTRYPOINT is set by current Claude Code versions too, and it is the only Claude marker MCP servers inherit, so it is not an "older versions" fallback. The Agent SDK comment still said Claude Code's shell tool sets CLAUDECODE; the rule keys on CLAUDE_CODE_CHILD_SESSION. Follow-up to review nits on heygen-com#7425. Comments only. Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

jrusso1020 and others added 2 commits November 9, 2026 02:47

docs(cli): trim comments and fix citations the comment gates flag

2ead55d

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

chore: raise the cli packed size budget to its measured size

5d2ee2e

main already sat at the edge of the 1% tolerance; the new agent rules put the packed CLI 842 bytes past it (8062664 vs 7982002 + 1%). Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

heygengenesis Bot approved these changes Nov 9, 2026

View reviewed changes

jrusso1020 added this pull request to the merge queue Nov 9, 2026

Merged via the queue into main with commit 7bf6599 Nov 9, 2026
93 of 94 checks passed

jrusso1020 deleted the agent-runtimes-lib-markers branch November 9, 2026 03:48

jrusso1020 mentioned this pull request Nov 9, 2026

docs(cli): correct Claude Code marker comments in agent detection #7426

Merged

This was referenced Nov 9, 2026

docs(changelog): weekly digest 2026-11-02-2026-11-09 #7453

Merged

chore(release): cut v0.9.27 #7464

Merged

fix(player): unload the video when its src is cleared #7466

Merged

Navigation Menu

Repository navigation

feat(cli): add agent markers other detectors use, verified upstream - #7425

feat(cli): add agent markers other detectors use, verified upstream#7425
jrusso1020 merged 3 commits into
mainfrom
agent-runtimes-lib-markers

jrusso1020 commented Nov 9, 2026 •

edited

Loading

Uh oh!

github-actions Bot commented Nov 9, 2026 •

edited

Loading

Uh oh!

heygengenesis Bot left a comment

Uh oh!

heygengenesis Bot Nov 9, 2026

Uh oh!

heygengenesis Bot Nov 9, 2026

Uh oh!

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

	// inherited CLAUDE_CODE_ENTRYPOINT (older versions). Not CLAUDECODE: its IDE
	// inherited CLAUDE_CODE_ENTRYPOINT (MCP servers, pre-2.1.172). Not CLAUDECODE: its IDE

	// Code's own shell tool (that path sets CLAUDECODE, caught above). The SDK
	// Code's own shell tool (that path sets CLAUDE_CODE_CHILD_SESSION, caught above). The SDK

Conversation

jrusso1020 commented Nov 9, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

What

Why

Related work

How

Test plan

Uh oh!

github-actions Bot commented Nov 9, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Edit accuracy: accurate 267 (base branch 267), smooth 113 of those

Uh oh!

heygengenesis Bot left a comment

Choose a reason for hiding this comment

Uh oh!

heygengenesis Bot Nov 9, 2026

Choose a reason for hiding this comment

Uh oh!

heygengenesis Bot Nov 9, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

jrusso1020 commented Nov 9, 2026 •

edited

Loading

github-actions Bot commented Nov 9, 2026 •

edited

Loading