feat: return native typed exec outcomes

[efecnc]

Status

Replacement review PR for #87. The original PR was merged accidentally and cannot be reopened by GitHub. PR #91 removed its changes from main; this PR restores the exact original diff for review and must remain open until explicitly approved for merge.

Summary

• derive exec success and failure directly from the child process status
• return NonZeroExit as an authoritative typed tool result while preserving legacy output text
• prevent successful command output that resembles an exit marker from being misclassified
• preserve large-output truncation, UTF-8 safety, and grep advisory behavior

Validation

• cargo fmt --all -- --check
• cargo clippy --all-targets --all-features -- -D warnings
• cargo test --all-features (440 passed, 0 failed, 6 ignored)
• git diff --check

Original PR: #87

[amazon-q-developer]

The changes successfully refactor the exec tool to derive execution success/failure directly from child process status rather than relying on parseable output markers. The implementation introduces the typed ToolResult with NonZeroExit error code while preserving backward compatibility.

Key improvements:

• Prevents command output from spoofing success/failure by separating process status from output content
• Introduces ShellExecOutcome struct and execute_with_approved_mutation_typed method for typed results
• Maintains existing execute method for backward compatibility
• Comprehensive test coverage including spoofing prevention tests

All validation passed (440 tests, clippy, fmt) and the code is ready for merge.

---

You can now have the agent implement changes and create commits directly on your pull request's source branch. Simply comment with /q followed by your request in natural language to ask the agent to make changes.

[efecnc]

Codex review

No blocking findings.

I traced both public execution paths: the reasoning loop reaches execute_with_approved_mutation_typed, where the child process status is authoritative, while the legacy execute() surface preserves its existing text contract. Successful output containing a forged Exit code: suffix stays successful, and genuine non-zero status retains its typed NonZeroExit root cause after advisory/truncation formatting. Unix edge tests, full tests, fmt, and clippy -D warnings are clean.

Review status: ready for human approval, but leave this PR open until explicitly authorized to merge.

[monatis]

lgtm