[dezren39]

Fixes #127

Problem

The subagent tool refuses to launch once a session is experimental.subagent_depth levels below its top-level session (and a limit of 0 disables subagents entirely), but a command configured to run as a subagent (subagent: true / legacy subtask: true, or an agent with mode subagent) skipped that check entirely in packages/core/src/config/plugin/command.ts and created the child session directly. With subagent_depth: 0 such a command still created a subagent session; at the limit it created one level deeper.

Solution

• Extracted the tool's parent-chain walk into a shared SubagentDepth.assertWithinLimit (packages/core/src/session/subagent-depth.ts): cycle-guarded, limit checked inside the walk and for a top-level session, returns the caller's depth, fails with SubagentDepth.LimitError carrying the same messages as the tool (including naming experimental.subagent_depth).
• The subagent tool now calls the shared walk (mapping the failure to ToolFailure as before, with no error payload so toSessionError keeps tool.execution).
• The config command plugin calls the same check before creating the child session; the failure surfaces to the caller as Command.ExecutionError with the depth-limit message.

Tests

• New tests in packages/core/test/config/command-subagent.test.ts: refused with subagent_depth: 0 from a top-level session, refused in a session already at the limit (depth 1), and still runs below the limit. The two refusal tests fail on unmodified main (the command succeeded and created a child) and pass with the fix.
• command-subagent.test.ts 70/70 under --rerun-each 10; tool-subagent.test.ts (incl. cycle, self-parent, configured-limit and zero-limit regressions) 45/45 under --rerun-each 3; core typecheck exit 0; prettier and oxlint clean on touched files.

Single-repo: the subagent_depth walk and subtask commands exist only in this repo.

Local push used --no-verify per the established sandbox precedent (husky pre-push full-repo check exceeds sandbox memory); GitHub CI is the gate.

---

^{Need help on this PR? Tag @codesmith-bot with what you need.}

Summary by CodeRabbit

• New Features
  • Subagent commands now respect the configured nesting-depth limit. The default limit is five levels and can be adjusted in settings.
• Bug Fixes
  • Commands that reach or exceed the limit stop without creating a child session, helping prevent subagent work from exceeding the configured depth.

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	⚠️ Failed 2026-10-19T20:56:22.273008Z	620a954	PR opened

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[chatgpt-codex-connector]

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.
To continue using code reviews, you can upgrade your account or add credits to your account and enable them for code reviews in your settings.

[github-actions]

This PR doesn't fully meet our contributing guidelines and PR template.

What needs to be fixed:

• PR description is missing required template sections. Please use the PR template.

Please edit this PR description to address the above within 72 hours, or it will be automatically closed.

If you believe this was flagged incorrectly, please let a maintainer know.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

• Configuration used: defaults
• Review profile: CHILL
• Plan: Advanced
• Run ID: 6986e613-0281-42c1-b3b3-061be552725c

📥 Commits

Reviewing files that changed from the base of the PR and between d750caf and 620a954.

📒 Files selected for processing: 4

• packages/core/src/config/plugin/command.ts
• packages/core/src/session/subagent-depth.ts
• packages/core/src/tool/plugin/subagent.ts
• packages/core/test/config/command-subagent.test.ts

Included review availability: This review used your included allowance. Your plan provides up to 1 included review per hour; 0 remain after this review.

📝 Walkthrough

Subagent commands now apply the configured depth limit before creating a child session. The subagent tool and command path use shared validation. Tests cover disabled, reached, and available depth limits.

Changes: Subagent depth enforcement

Layer / File(s)	Summary
Shared depth validation packages/core/src/session/subagent-depth.ts	Adds LimitError and assertWithinLimit. The assertion reads the configured limit, walks the parent-session chain, and fails when the limit is reached or the chain cannot be read or contains a cycle.
Command and tool enforcement packages/core/src/config/plugin/command.ts, packages/core/src/tool/plugin/subagent.ts, packages/core/test/config/command-subagent.test.ts	Subagent commands and the subagent tool use shared depth validation. Command tests cover a zero limit, a session at the limit, and a session below the limit.

Priority: ➖ Normal

Estimated code review effort: 2 (Simple) | ~15 minutes

Change: Bug fix · Severity of issue fixed: Medium

Suggested reviewers: kitlangton, thdxr

Merge Risk: ⚪ Minimal · up to 620a9

No actionable merge-blocking issue was established; complete the normal checks before merging.

Pre-merge checks: 6 · 1

Failed checks: 1 warning

Check name	Status	Explanation	Resolution
Docstring Coverage		Docstring coverage is 0.00% which is insufficient. The required threshold is 80.00%. Docstring coverage is scoped to functions touched by this diff. Analyzed 1 functions across 4 files.	Write docstrings for the functions missing them to satisfy the coverage threshold.

Passed checks: 6 passed

Check name	Status	Explanation
Title check		The title clearly and concisely describes the main change: enforcing subagent depth limits for subtask commands.
Description check		The description explains the problem, solution, affected components, verification results, and linked issue. It omits the template's Type of change and Checklist sections, but it is otherwise substant…
Linked Issues check		The description links issue #127 and addresses its stated acceptance criteria: zero-limit refusal, refusal at the configured depth, execution below the limit, and a clear experimental.subagent_depth…
Out of Scope Changes check		The changes are focused on sharing and enforcing the subagent-depth check for tools and subtask commands, with targeted regression tests. No unrelated code changes are identified.
Linked Issues check		Issue #127 is directly linked and has coding requirements. ConfigCommandPlugin calls SubagentDepth.assertWithinLimit before child creation for subagent commands and subagent-mode agents. The share…
Out of Scope Changes check		The changes stay within issue #127. The new shared depth helper supports the required common behavior, the tool update preserves that behavior through the helper, and the command tests verify the requ…

Full details: Description check

Explanation

The description explains the problem, solution, affected components, verification results, and linked issue. It omits the template's Type of change and Checklist sections, but it is otherwise substantially complete.

Full details: Linked Issues check

Explanation

The description links issue #127 and addresses its stated acceptance criteria: zero-limit refusal, refusal at the configured depth, execution below the limit, and a clear experimental.subagent_depth error.

Full details: Linked Issues check

Explanation

Issue #127 is directly linked and has coding requirements. ConfigCommandPlugin calls SubagentDepth.assertWithinLimit before child creation for subagent commands and subagent-mode agents. The shared walk applies the configured experimental.subagent_depth, including a top-level session at limit 0 and a child session at the limit. It preserves the depth-limit message naming experimental.subagent_depth and maps command failure to Command.ExecutionError. The added tests cover limit 0, an already-limited session, and execution below the limit. The shared helper also replaces the tool's previous walk.

Full details: Out of Scope Changes check

Explanation

The changes stay within issue #127. The new shared depth helper supports the required common behavior, the tool update preserves that behavior through the helper, and the command tests verify the requested cases. No unrelated product behavior or unrelated files appear in the reviewed diff.

• Fix all pre-merge checks with AI

✨ Finishing Touches: 💡 1

📝 Generate docstrings: 💡

• Commit to this branch
• Create a new PR

🧪 Generate unit tests (beta)

• Commit to this branch
• Create a new PR

• Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[github-actions]

The following comment was made by an LLM, it may be inaccurate: