feat(core): enhance shell command validation and add core tools allowlist

[galz10]

Summary

This PR enhances the policy engine's shell command validation by implementing recursive checking for sub-commands, substitutions, and subshells. It also introduces a tools.core setting to allow users to explicitly allowlist specific core tools with high precision.

Details

• Recursive Validation: The PolicyEngine now utilizes parseCommandDetails from shell-utils to identify all nested parts of a shell command (substitutions, subshells, piped commands). Each part is checked against the policy rules.
• Core Tools Allowlist: Added settings.tools.core support. This allows defining an explicit list of allowed core tool invocations (e.g., run_shell_command(ls)).
• Shell Wrapper Handling: Added stripShellWrapper to handle commands executed via bash -c and similar wrappers recursively.
• Improved Heuristics: Refactored the check logic to be more robust against complex chained commands and redirections.

Related Issues

N/A

How to Validate

Run the new regression test suites:

1. npm test packages/core/src/policy/core-tools-mapping.test.ts
2. npm test packages/core/src/policy/shell-safety-regression.test.ts
3. npm test packages/core/src/policy/shell-substitution.test.ts

Pre-Merge Checklist

• Updated relevant documentation and README (if needed)
• Added/updated tests (if needed)
• Noted breaking changes (if any)
• Validated on MacOS

[github-actions]

Size Change: +2.53 kB (+0.01%)

Total Size: 33.7 MB

Filename	Size	Change
./bundle/chunk-47IBHWBB.js	0 B	-14.6 MB (removed)	🏆
./bundle/chunk-AVLKHPC5.js	0 B	-669 kB (removed)	🏆
./bundle/chunk-GEHXKXQJ.js	0 B	-3.43 kB (removed)	🏆
./bundle/chunk-UZ5TYF7Z.js	0 B	-2.73 MB (removed)	🏆
./bundle/chunk-VSIOCFQH.js	0 B	-3.8 kB (removed)	🏆
./bundle/chunk-ZZQWEWOR.js	0 B	-49.2 kB (removed)	🏆
./bundle/core-DF3YXFQ2.js	0 B	-47.4 kB (removed)	🏆
./bundle/devtoolsService-EZXYI5WH.js	0 B	-27.8 kB (removed)	🏆
./bundle/gemini-XL3WLALG.js	0 B	-578 kB (removed)	🏆
./bundle/interactiveCli-BNUZ2WFS.js	0 B	-1.29 MB (removed)	🏆
./bundle/liteRtServerManager-YEOWMMKG.js	0 B	-2.08 kB (removed)	🏆
./bundle/oauth2-provider-SXZTOYIT.js	0 B	-9.16 kB (removed)	🏆
./bundle/chunk-44SZCZMU.js	3.43 kB	+3.43 kB (new file)	🆕
./bundle/chunk-4QUNGO5R.js	3.8 kB	+3.8 kB (new file)	🆕
./bundle/chunk-5A3CYYIV.js	49.2 kB	+49.2 kB (new file)	🆕
./bundle/chunk-5IJP6J5Q.js	669 kB	+669 kB (new file)	🆕
./bundle/chunk-6DA56GQW.js	14.6 MB	+14.6 MB (new file)	🆕
./bundle/chunk-6I46UJMF.js	2.73 MB	+2.73 MB (new file)	🆕
./bundle/core-5G6EFRH2.js	47.5 kB	+47.5 kB (new file)	🆕
./bundle/devtoolsService-ACRICRWS.js	27.8 kB	+27.8 kB (new file)	🆕
./bundle/gemini-QL4SO2IH.js	578 kB	+578 kB (new file)	🆕
./bundle/interactiveCli-EKSFLAZB.js	1.29 MB	+1.29 MB (new file)	🆕
./bundle/liteRtServerManager-DPJADFO6.js	2.08 kB	+2.08 kB (new file)	🆕
./bundle/oauth2-provider-4OJV74JW.js	9.16 kB	+9.16 kB (new file)	🆕

ℹ️ View Unchanged

Filename	Size	Change
./bundle/bundled/third_party/index.js	8 MB	0 B
./bundle/chunk-34MYV7JD.js	2.45 kB	0 B
./bundle/chunk-5AUYMPVF.js	858 B	0 B
./bundle/chunk-5PS3AYFU.js	1.18 kB	0 B
./bundle/chunk-664ZODQF.js	124 kB	0 B
./bundle/chunk-DAHVX5MI.js	206 kB	0 B
./bundle/chunk-IUUIT4SU.js	56.5 kB	0 B
./bundle/chunk-RJTRUG2J.js	39.8 kB	0 B
./bundle/chunk-UHHRGNIO.js	1.97 MB	0 B
./bundle/cleanup-EMA3ZJSE.js	0 B	-932 B (removed)	🏆
./bundle/devtools-36NN55EP.js	696 kB	0 B
./bundle/dist-T73EYRDX.js	356 B	0 B
./bundle/events-XB7DADIJ.js	418 B	0 B
./bundle/examples/hooks/scripts/on-start.js	188 B	0 B
./bundle/examples/mcp-server/example.js	1.43 kB	0 B
./bundle/gemini.js	4.97 kB	0 B
./bundle/getMachineId-bsd-TXG52NKR.js	1.55 kB	0 B
./bundle/getMachineId-darwin-7OE4DDZ6.js	1.55 kB	0 B
./bundle/getMachineId-linux-SHIFKOOX.js	1.34 kB	0 B
./bundle/getMachineId-unsupported-5U5DOEYY.js	1.06 kB	0 B
./bundle/getMachineId-win-6KLLGOI4.js	1.72 kB	0 B
./bundle/memoryDiscovery-CZURD35O.js	980 B	0 B
./bundle/multipart-parser-KPBZEGQU.js	11.7 kB	0 B
./bundle/node_modules/@google/gemini-cli-devtools/dist/client/main.js	222 kB	0 B
./bundle/node_modules/@google/gemini-cli-devtools/dist/src/_client-assets.js	229 kB	0 B
./bundle/node_modules/@google/gemini-cli-devtools/dist/src/index.js	13.4 kB	0 B
./bundle/node_modules/@google/gemini-cli-devtools/dist/src/types.js	132 B	0 B
./bundle/sandbox-macos-permissive-open.sb	890 B	0 B
./bundle/sandbox-macos-permissive-proxied.sb	1.31 kB	0 B
./bundle/sandbox-macos-restrictive-open.sb	3.36 kB	0 B
./bundle/sandbox-macos-restrictive-proxied.sb	3.56 kB	0 B
./bundle/sandbox-macos-strict-open.sb	4.82 kB	0 B
./bundle/sandbox-macos-strict-proxied.sb	5.02 kB	0 B
./bundle/src-QVCVGIUX.js	47 kB	0 B
./bundle/start-TYOHETGR.js	0 B	-622 B (removed)	🏆
./bundle/tree-sitter-7U6MW5PS.js	274 kB	0 B
./bundle/tree-sitter-bash-34ZGLXVX.js	1.84 MB	0 B
./bundle/cleanup-QZESAQ2R.js	932 B	+932 B (new file)	🆕
./bundle/start-ALP62AHI.js	622 B	+622 B (new file)	🆕

_{compressed-size-action}

[gemini-code-assist]

Summary of Changes

Hello, I'm Gemini Code Assist¹! I'm currently reviewing this pull request and will post my feedback shortly. In the meantime, here's a summary to help you and other reviewers quickly get up to speed!

This pull request significantly strengthens the security posture of the shell command validation system. By enabling recursive parsing and validation of nested shell structures, the engine can now detect and restrict potentially dangerous commands that were previously bypassed. Additionally, the introduction of a dedicated core tools allowlist provides users with finer-grained control over essential tool execution, improving both security and usability.

Highlights

• Recursive Shell Validation: Enhanced the PolicyEngine to recursively validate nested shell commands, including sub-commands, substitutions, and subshells, by utilizing the updated shell-utils parser.
• Core Tools Allowlist: Introduced a new settings.tools.core configuration that allows users to explicitly define permitted core tool invocations with high precision.
• Shell Wrapper Handling: Implemented recursive stripping of shell wrappers (e.g., bash -c) to ensure that commands executed within wrappers are correctly validated against policy rules.
• Robust Heuristics: Refactored the policy check logic to be more resilient against complex chained commands and redirections, ensuring consistent security enforcement.

Using Gemini Code Assist

The full guide for Gemini Code Assist can be found on our documentation page, here are some quick tips.

Invoking Gemini

You can request assistance from Gemini at any point by creating a comment using either /gemini <command> or @gemini-code-assist <command>. Below is a summary of the supported commands on the current page.

Feature	Command	Description
Code Review	/gemini review	Performs a code review for the current pull request in its current state.
Pull Request Summary	/gemini summary	Provides a summary of the current pull request in its current state.
Comment	@gemini-code-assist	Responds in comments when explicitly tagged, both in pull request comments and review comments.
Help	/gemini help	Displays a list of available commands.

Customization

To customize Gemini Code Assist for GitHub experience, repository maintainers can create a configuration file and/or provide a custom code review style guide (such as PEP-8 for Python) by creating and adding files to a .gemini/ folder in the base of the repository. Detailed instructions can be found here.

Limitations & Feedback

Gemini Code Assist may make mistakes. Please leave feedback on any instances where its feedback is incorrect or counter productive. You can react with 👍 and 👎 on @gemini-code-assist comments. If you're interested in giving your feedback about your experience with Gemini Code Assist for GitHub and other Google products, sign up here.

Footnotes

1. Review the Privacy Notices, Generative AI Prohibited Use Policy, Terms of Service, and learn how to configure Gemini Code Assist in GitHub here. Gemini can make mistakes, so double check it and use code with caution. ↩

[gemini-code-assist]

Code Review

This pull request introduces a core tools allowlist feature and significantly enhances shell command safety validation within the policy engine. Key changes include refactoring the policy configuration to support a new tools.core setting, implementing recursive validation for shell wrappers and command substitutions, and adding comprehensive regression tests for various shell constructs like pipes and redirections. Feedback was provided regarding the default decision for the core tools allowlist; currently, it defaults to DENY, but per project guidelines, it should default to ALLOW until a formal approval mechanism is integrated to avoid being overly restrictive.

[gemini-code-assist]

The implementation currently defaults to a DENY decision when tools.core is configured. Based on the repository rules, the default policy decision for tool calls should be ALLOW until an approval mechanism is in place, as a DENY default is considered too restrictive. Please update the decision to ALLOW and ensure the priority is correctly set to avoid unintentionally overriding other policies.

Suggested change

	rules.push({
	toolName: '*',
	decision: PolicyDecision.DENY,
	priority: CORE_TOOLS_FLAG_PRIORITY - 0.01,
	source: 'Settings (Core Tools Allowlist Enforcement)',
	});
	rules.push({
	toolName: '*',
	decision: PolicyDecision.ALLOW,
	priority: USER_POLICY_TIER + 0.05,
	source: 'Settings (Core Tools Allowlist Enforcement)',
	});

References

1. The default policy decision for tool calls should be ALLOW until a mechanism for wiring up approvals is integrated, as a DENY default would be too restrictive without an approval flow.

[DavidAPierce]

This comment seems to be ignoring the secure by default methodology being implemented when core tools are restricted. I don't think that this is a valid concern from the bot's part, but if you think this is correct, feel free to address the above comment.

[DavidAPierce]

LGTM. I also ran it through my CLI review process locally, and it had some nice things to say, so here they are:

Summary
The changes successfully address complex shell nesting scenarios, including command substitutions and shell wrappers (e.g., bash -c). The implementation leverages the Tree-sitter parser for precise command decomposition
and applies a recursive checking strategy that is both secure and efficient.

Findings

Improvements & Robustness

• Recursive Validation Depth: I verified that the recursive logic correctly handles nested wrappers and substitutions (e.g., bash -c "echo $(ls)"). Even when outer parsers cannot see inside single-quoted strings, the
  stripShellWrapper logic triggers a secondary parse on the inner command, ensuring no hidden commands bypass the policy engine.
• Tree-sitter Integration: Transitioning to parseCommandDetails (using Tree-sitter) provides much higher fidelity than simple regex or string splitting, correctly identifying subshells and substitutions as distinct
  components.
• Core Tools Allowlist: The new settings.tools.core configuration provides a powerful way for users to enforce a "Least Privilege" model for shell access.

Discussion Points

• Global Impact of tools.core: When settings.tools.core is defined, the engine adds a global DENY rule (toolName: '*') at priority CORE_TOOLS_FLAG_PRIORITY - 0.01.
  • Observation: This priority (e.g., 4.24) is higher than the default priorities for MCP servers (4.1 and 4.2). Consequently, enabling a core tools allowlist will effectively block all MCP tools unless they are also
    explicitly allowlisted.
  • Recommendation: This "Strict Mode" behavior is highly secure but should be explicitly documented to ensure users understand that tools.core is a global allowlist enforcement mechanism, not just a restriction on
    core tools.

Nitpicks

• REDIRECTION_NAMES naming: Adding command substitution and subshell to REDIRECTION_NAMES in shell-utils.ts is effective for the current logic (which skips these nodes in certain loops), but the name REDIRECTION_NAMES
  is now slightly technically inaccurate as it includes non-redirection shell constructs. A rename to something like SKIPPABLE_STRUCTURAL_NODES might be clearer in the future, but this is a minor point.

Conclusion

Status: Approved

The implementation is technically sound, well-tested, and provides a significant security improvement. The recursive stripping and parsing strategy is particularly impressive for its thoroughness.

[Abhijit-2592]

LGTM

[github-actions]

✅ 57 tests passed successfully on gemini-3-flash-preview.

🧠 Model Steering Guidance

This PR modifies files that affect the model's behavior (prompts, tools, or instructions).

• 🚀 Maintainer Reminder: Please ensure that these changes do not regress results on benchmark evals before merging.

---

This is an automated guidance message triggered by steering logic signatures.