feat(agents): add Undefined self code inspection agent#74
Merged
Conversation
Co-authored-by: GPT-5 Codex <noreply@openai.com>
Co-authored-by: GPT-5 Codex <noreply@openai.com>
|
You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard. |
Contributor
|
No actionable comments were generated in the recent review. 🎉 ℹ️ Recent review info⚙️ Run configurationConfiguration used: Path: .coderabbit.yaml Review profile: CHILL Plan: Pro Plus Run ID: 📒 Files selected for processing (6)
🚧 Files skipped from review as they are similar to previous changes (5)
📝 WalkthroughWalkthroughA new Changesundefined_self_code_agent: read-only code inspection agent
Sequence Diagram(s)sequenceDiagram
participant LLM as Orchestrator LLM
participant AgentHandler as undefined_self_code_agent/handler.py
participant ToolHandler as Tool Handler<br/>read_file/list_directory/glob/search
participant Shared as _shared.py
participant FS as Filesystem
LLM->>AgentHandler: execute({prompt: "..."}, context)
AgentHandler->>AgentHandler: run_agent_with_tools(...)
AgentHandler->>ToolHandler: call tool(args, context)
ToolHandler->>ToolHandler: validate + parse args
ToolHandler->>Shared: resolve_allowed_path or<br/>resolve_search_root
Shared->>Shared: find_repo_root(context)
Shared->>Shared: is_allowed_path(resolved_path)
alt path outside allowlist or traversal detected
Shared-->>ToolHandler: PermissionError
ToolHandler-->>AgentHandler: 权限不足：{allowed_roots_text()}
else path allowed
Shared-->>ToolHandler: ResolvedPath
alt read_file
ToolHandler->>FS: check exists + is_file
ToolHandler->>Shared: read_text_file in thread
Shared-->>ToolHandler: (text, truncated, size)
ToolHandler-->>AgentHandler: file header + content
else list_directory
ToolHandler->>Shared: list_allowed_directory_entries
Shared-->>ToolHandler: [(rel_path, is_dir), ...]
ToolHandler-->>AgentHandler: 📁📄 sorted list
else glob
ToolHandler->>Shared: collect_allowed_glob_matches in thread
Shared-->>ToolHandler: [rel_path, ...]
ToolHandler-->>AgentHandler: newline-joined paths
else search_file_content
ToolHandler->>Shared: iter_allowed_files + compile_pattern
ToolHandler->>Shared: read_text_file per file
Shared-->>ToolHandler: decoded text + line numbers
ToolHandler-->>AgentHandler: path:line:content matches
end
end
AgentHandler-->>LLM: agent response
Estimated code review effort🎯 4 (Complex) | ⏱️ ~60 minutes Poem
🚥 Pre-merge checks | ✅ 4 | ❌ 1❌ Failed checks (1 warning)
✅ Passed checks (4 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches📝 Generate docstrings
🧪 Generate unit tests (beta)
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. Comment |
Owner
Author
|
@coderabbitai review |
Contributor
✅ Action performedReview finished.
|
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
Contributor
There was a problem hiding this comment.
Actionable comments posted: 10
🧹 Nitpick comments (1)
tests/test_naga_code_analysis_agent.py (1)
14-14: ⚡ Quick winAvoid pinning this test to an exact submodule hash marker.
Asserting
"eb71318"makes the test fail on normal NagaAgent syncs even when behavior is correct. Prefer stable capability/layout markers only.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@tests/test_naga_code_analysis_agent.py` at line 14, The assertion checking for the exact submodule hash "eb71318" is too brittle and causes test failures during normal submodule syncs even when the behavior is correct. Replace this exact hash check with stable capability or layout markers that validate the correct behavior without depending on specific commit hashes.
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@ARCHITECTURE.md`:
- Line 120: The A_Self node documentation indicates the
undefined_self_code_agent has 4 tools (as shown by "4个工具"), but the bullet list
under it only documents 3 tools: read_file / glob and search_file_content. Add a
new bullet point to include list_directory in the tool list to accurately
document all 4 tools that the agent actually exposes.
In `@res/prompts/undefined_nagaagent.xml`:
- Line 889: In the rule describing undefined_self_code_agent capabilities, there
is a duplicated "只" character in the phrase "只能只读查阅". Remove one instance of the
"只" character to normalize the phrase and ensure it reads naturally without
redundancy. The phrase should convey that the agent can only read and consult
the code, not write or execute commands.
In `@res/prompts/undefined.xml`:
- Line 838: In the rule describing undefined_self_code_agent limitations, remove
the duplicate character "只" from the phrase `只能只读查阅`. Replace `只能只读查阅` with
either `只能查阅` (simpler) or `仅可只读查阅` (preserving the read-only intent) to make
the sentence grammatically correct and easier to read. The phrase appears in the
constraint text that describes what undefined_self_code_agent cannot do.
In
`@src/Undefined/skills/agents/naga_code_analysis_agent/tools/read_naga_intro/handler.py`:
- Around line 6-71: Update the outdated startup and build command references in
the embedded intro text within the handler.py file (around lines 93–98). Replace
the non-existent entry points `python apiserver/start_server.py api` and
`uvicorn agentserver.agent_server:app` with the correct unified entry point
`python main.py` or `python -m Undefined`. Update the build script reference
from `python scripts/build-win.py` to point to the correct build script that
exists in the `scripts/` directory (such as `build_native_apps.py`). Update the
documentation reference from `docs/build-windows.md` to `docs/build.md`. Ensure
all command examples and file paths in the intro match the actual current
repository structure.
In `@src/Undefined/skills/agents/undefined_self_code_agent/tools/_shared.py`:
- Around line 165-190: The function iter_allowed_files uses blocking filesystem
operations (exists, is_file, is_dir, rglob) that can stall the event loop in
async contexts. Replace these direct blocking calls with async-safe helpers from
src/Undefined/utils/io.py. Convert iter_allowed_files from a synchronous
generator to an async generator that uses asyncio.to_thread() to wrap blocking
filesystem operations, ensuring file existence checks, directory traversals, and
path validation are properly handled through the project's I/O layer instead of
direct Path method calls.
In `@src/Undefined/skills/agents/undefined_self_code_agent/tools/glob/handler.py`:
- Around line 44-59: The synchronous filesystem traversal operations in the loop
spanning lines 44-59 (iterating through search_roots, calling glob(), is_file(),
is_allowed_path(), format_relative(), and managing matches) are executing
directly on the event loop and blocking concurrent work. Extract this entire
disk traversal logic into a separate synchronous helper function, then call that
function using asyncio.to_thread() from the async handler to offload the
blocking I/O operations to a thread pool, following the coding guidelines
requirement to use asyncio.to_thread for disk read operations.
In
`@src/Undefined/skills/agents/undefined_self_code_agent/tools/list_directory/handler.py`:
- Around line 40-42: The execute method is performing blocking synchronous
filesystem operations (iterdir() calls and is_dir() checks) directly in the
async coroutine context on lines 40-42 and 57-59, which blocks the event loop.
Create utility functions in the utils/io.py module to handle directory scanning
and file type checking, then import and call these utilities in the execute
method using asyncio.to_thread to execute them asynchronously. This ensures all
disk I/O operations are wrapped appropriately and non-blocking within the async
context.
In
`@src/Undefined/skills/agents/undefined_self_code_agent/tools/read_file/handler.py`:
- Around line 51-60: The range_header construction on line 59 produces an
invalid format like "行 1-0/0" when the file is empty. Add a conditional check
after calculating end_line to detect when the file is empty (when total_lines is
0 or selected is empty) and set range_header to an explicit empty-range header
in that case, otherwise use the existing range_header format. This ensures the
header accurately represents empty files instead of emitting invalid line
ranges.
In
`@src/Undefined/skills/agents/undefined_self_code_agent/tools/search_file_content/handler.py`:
- Around line 57-76: The code is performing synchronous blocking file operations
directly in an async context by eagerly materializing all files with
list(iter_allowed_files(...)) and calling read_text_file() synchronously.
Convert the function to be async and wrap the blocking IO operations with
asyncio.to_thread() when calling read_text_file() and any other disk operations.
Remove the list() call around iter_allowed_files() to make file enumeration
lazy, allowing the event loop to continue processing while file operations
complete. This prevents blocking the event loop and reduces memory usage on
larger repositories.
In `@tests/test_undefined_self_code_agent.py`:
- Around line 46-71: Replace all direct Path.write_text() calls throughout the
test setup with the project's async-safe I/O utilities from
src/Undefined/utils/io.py. Each line creating files using the pattern (root /
"...").write_text() should be refactored to use the appropriate utility function
from the io module, ensuring all disk writes are atomic and async-safe as
required by the project's coding guidelines. This applies to all file creation
operations including pyproject.toml, README.md, CHANGELOG.md, ARCHITECTURE.md,
config.toml.example, main.py, .hidden.py, tool.py, test_main.py, prompt.txt,
usage.md, App.tsx, secret.txt, run.log, and .env files.
---
Nitpick comments:
In `@tests/test_naga_code_analysis_agent.py`:
- Line 14: The assertion checking for the exact submodule hash "eb71318" is too
brittle and causes test failures during normal submodule syncs even when the
behavior is correct. Replace this exact hash check with stable capability or
layout markers that validate the correct behavior without depending on specific
commit hashes.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro Plus
Run ID: 9e99344b-ef36-4839-b6d6-88b556c8a382
📒 Files selected for processing (28)
ARCHITECTURE.mdCHANGELOG.mdREADME.mdcode/NagaAgentdocs/usage.mdres/prompts/undefined.xmlres/prompts/undefined_nagaagent.xmlsrc/Undefined/skills/agents/README.mdsrc/Undefined/skills/agents/naga_code_analysis_agent/tools/read_naga_intro/handler.pysrc/Undefined/skills/agents/undefined_self_code_agent/README.mdsrc/Undefined/skills/agents/undefined_self_code_agent/__init__.pysrc/Undefined/skills/agents/undefined_self_code_agent/config.jsonsrc/Undefined/skills/agents/undefined_self_code_agent/handler.pysrc/Undefined/skills/agents/undefined_self_code_agent/intro.mdsrc/Undefined/skills/agents/undefined_self_code_agent/prompt.mdsrc/Undefined/skills/agents/undefined_self_code_agent/tools/__init__.pysrc/Undefined/skills/agents/undefined_self_code_agent/tools/_shared.pysrc/Undefined/skills/agents/undefined_self_code_agent/tools/glob/config.jsonsrc/Undefined/skills/agents/undefined_self_code_agent/tools/glob/handler.pysrc/Undefined/skills/agents/undefined_self_code_agent/tools/list_directory/config.jsonsrc/Undefined/skills/agents/undefined_self_code_agent/tools/list_directory/handler.pysrc/Undefined/skills/agents/undefined_self_code_agent/tools/read_file/config.jsonsrc/Undefined/skills/agents/undefined_self_code_agent/tools/read_file/handler.pysrc/Undefined/skills/agents/undefined_self_code_agent/tools/search_file_content/config.jsonsrc/Undefined/skills/agents/undefined_self_code_agent/tools/search_file_content/handler.pytests/test_naga_code_analysis_agent.pytests/test_system_prompt_constraints.pytests/test_undefined_self_code_agent.py
Co-authored-by: GPT-5 Codex <noreply@openai.com>
Contributor
There was a problem hiding this comment.
Actionable comments posted: 3
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@src/Undefined/skills/agents/undefined_self_code_agent/tools/_shared.py`:
- Around line 207-217: The glob operation in the loop starting with `for root in
roots:` traverses the entire repository including disallowed directories before
filtering occurs, and patterns containing `..` can force traversal outside the
search root. Before calling `root.glob(pattern)`, validate that the pattern does
not contain `..` or absolute path components that could enable traversal outside
allowed roots. Additionally, add explicit logic to handle allowed root files
separately and ensure that glob operations are constrained to only search within
the approved roots rather than relying on post-glob filtering with
`is_allowed_path()` to exclude unwanted directories.
In `@src/Undefined/skills/agents/undefined_self_code_agent/tools/glob/handler.py`:
- Line 13: The handler.py file directly imports from Undefined.utils (the
async_io module), which violates the skills handler boundary by importing from
outside the skills/ directory. Remove the direct import statement "from
Undefined.utils import io as async_io" and instead either inject the async_io
dependency through the execution context parameter passed to the handler, or
expose it via _shared.py and import from there instead. This ensures
dependencies are properly passed through the context rather than directly
importing from repo-local modules outside the skills/ directory.
In
`@src/Undefined/skills/agents/undefined_self_code_agent/tools/list_directory/handler.py`:
- Line 13: The handler at list_directory/handler.py violates the skills module
boundary by directly importing from Undefined.utils (line 13: from
Undefined.utils import io as async_io). Remove this direct import statement and
instead inject the async_io dependency through the execution context that is
passed to the handler function. Update all usages of async_io within the handler
to reference it from the injected context parameter rather than the module
import.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro Plus
Run ID: 75ec38e4-5b4d-4cd7-b020-39fa5c122da8
📒 Files selected for processing (13)
ARCHITECTURE.mdres/prompts/undefined.xmlres/prompts/undefined_nagaagent.xmlsrc/Undefined/skills/agents/naga_code_analysis_agent/tools/read_naga_intro/handler.pysrc/Undefined/skills/agents/undefined_self_code_agent/tools/_shared.pysrc/Undefined/skills/agents/undefined_self_code_agent/tools/glob/handler.pysrc/Undefined/skills/agents/undefined_self_code_agent/tools/list_directory/handler.pysrc/Undefined/skills/agents/undefined_self_code_agent/tools/read_file/handler.pysrc/Undefined/skills/agents/undefined_self_code_agent/tools/search_file_content/handler.pysrc/Undefined/utils/io.pytests/test_naga_code_analysis_agent.pytests/test_system_prompt_constraints.pytests/test_undefined_self_code_agent.py
✅ Files skipped from review due to trivial changes (2)
- src/Undefined/skills/agents/naga_code_analysis_agent/tools/read_naga_intro/handler.py
- ARCHITECTURE.md
🚧 Files skipped from review as they are similar to previous changes (6)
- res/prompts/undefined.xml
- src/Undefined/skills/agents/undefined_self_code_agent/tools/search_file_content/handler.py
- tests/test_system_prompt_constraints.py
- res/prompts/undefined_nagaagent.xml
- tests/test_undefined_self_code_agent.py
- src/Undefined/skills/agents/undefined_self_code_agent/tools/read_file/handler.py
Co-authored-by: GPT-5 Codex <noreply@openai.com>
Owner
Author
|
@coderabbitai review |
Contributor
✅ Action performedReview finished.
|
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Tests
Summary by CodeRabbit
Release Notes
New Features
undefined_self_code_agentfor read-only lookup of Undefined’s own source/tests/docs/resources/scripts and app implementation, restricted to an allowlisted set of directories and root files.Documentation
skills/agents/.Tests
Chores