feat: add skills feature to advanced agent with workspace to load skills#977
Open
mariadhakalUipath wants to merge 1 commit into
Open
feat: add skills feature to advanced agent with workspace to load skills#977mariadhakalUipath wants to merge 1 commit into
mariadhakalUipath wants to merge 1 commit into
Conversation
There was a problem hiding this comment.
Pull request overview
Adds opt-in “skills” support to deepagents-backed advanced agents, including workspace-mounted skill sources and auto-injected sandboxed workspace tools (CLI runner, audit log reader, and scaffold verifier) when running on a FilesystemBackend.
Changes:
- Adds a
skills: Sequence[str] | Noneparameter tocreate_advanced_agent/create_advanced_agent_graphand wires it into deepagents’ SkillsMiddleware with workspace-source prepending/dedup. - Introduces three workspace-bound internal tools (
uipath_cli,uipath_cli_usage,uipath_verify_files) plus timeout/audit/denylist and required-files validation logic. - Adds focused test coverage for skills opt-in behavior, tool injection gating, CLI sandboxing/audit behavior, usage reading, and file verification.
Reviewed changes
Copilot reviewed 10 out of 10 changed files in this pull request and generated 5 comments.
Show a summary per file
| File | Description |
|---|---|
| src/uipath_langchain/agent/advanced/agent.py | Wires skills into deepagents and injects workspace tools when skills are active on a filesystem backend. |
| src/uipath_langchain/agent/advanced/utils.py | Defines skills workspace layout constants (SKILLS_DIR_NAME, SKILLS_VIRTUAL_PATH). |
| src/uipath_langchain/agent/advanced/init.py | Exports skills-related constants from the advanced agent package. |
| src/uipath_langchain/agent/tools/internal_tools/uipath_cli_tool.py | Implements the sandboxed CLI runner, audit-log usage reader, and required-files verifier tools. |
| src/uipath_langchain/agent/tools/internal_tools/init.py | Exposes the new internal tool constructors/utilities from the internal tools package. |
| tests/agent/advanced/test_skills_injection.py | Tests skills opt-in gating and workspace source prepending/dedup behavior. |
| tests/agent/advanced/test_cli_tool_injection.py | Tests conditional injection of the three workspace tools based on skills + filesystem backend. |
| tests/agent/tools/internal_tools/test_uipath_cli_tool.py | Tests CLI sandboxing: binary allowlist, subdir jailing, timeout resolution, denylist, and auditing. |
| tests/agent/tools/internal_tools/test_cli_usage_tool.py | Tests reading and summarizing CLI audit log records (including malformed line handling). |
| tests/agent/tools/internal_tools/test_verify_files_tool.py | Tests file verification behavior including glob patterns, unknown types, and workspace escape prevention. |
Comment on lines
86
to
88
| memory=list(memory) or None, | ||
| skills=list(skills) if skills is not None else None, | ||
| ) |
Comment on lines
+152
to
+155
| workspace_path = Path(workspace).expanduser().resolve() | ||
| workspace_path.mkdir(parents=True, exist_ok=True) | ||
| audit_path = Path(audit_log or (workspace_path / AUDIT_LOG_FILENAME)).expanduser() | ||
| denied = frozenset( |
Comment on lines
+264
to
+272
| return finish( | ||
| "ok" if proc.returncode == 0 else "nonzero_exit", | ||
| f"exit_code: {proc.returncode}\n--- stdout ---\n{proc.stdout}\n--- stderr ---\n{proc.stderr}", | ||
| exit_code=proc.returncode, | ||
| stdout=proc.stdout[:MAX_LOG_BYTES], | ||
| stderr=proc.stderr[:MAX_LOG_BYTES], | ||
| stdout_truncated=len(proc.stdout) > MAX_LOG_BYTES, | ||
| stderr_truncated=len(proc.stderr) > MAX_LOG_BYTES, | ||
| ) |
Comment on lines
+281
to
+294
| def _read_audit_records(audit_path: Path) -> list[dict[str, Any]]: | ||
| """Return audit records of cli commands execution in invocation order; empty if the log is absent.""" | ||
| if not audit_path.is_file(): | ||
| return [] | ||
| records: list[dict[str, Any]] = [] | ||
| for raw in audit_path.read_text(encoding="utf-8").splitlines(): | ||
| line = raw.strip() | ||
| if not line: | ||
| continue | ||
| try: | ||
| records.append(json.loads(line)) | ||
| except json.JSONDecodeError: | ||
| continue | ||
| return records |
Comment on lines
+331
to
+335
| def _file_matches(project_dir: Path, pattern: str) -> bool: | ||
| """True if ``pattern`` (literal name or glob) matches a file in the dir root.""" | ||
| if any(ch in pattern for ch in "*?["): | ||
| return any(project_dir.glob(pattern)) | ||
| return (project_dir / pattern).is_file() |
mariadhakalUipath
force-pushed
the
feat/uipath-skills-for-advanced-agents
branch
2 times, most recently
from
e5d2b48 to
5cee6e8
Compare
Comment on lines
+118
to
+123
| binary = os.path.basename(argv[0]) | ||
| if binary not in ALLOWED_BINARIES: | ||
| return ( | ||
| f"ERROR: '{binary}' is not allowed. Only these binaries are " | ||
| f"permitted: {sorted(ALLOWED_BINARIES)}" | ||
| ) |
Comment on lines
+59
to
+60
| memory=list(memory) or None, | ||
| skills=list(skills) if skills is not None else None, |
| backend=backend, | ||
| response_format=response_format, | ||
| memory=list(memory) or None, | ||
| skills=list(skills) if skills is not None else None, |
There was a problem hiding this comment.
because of the default None in the function call this could be simply skills = list(skills)
Author
There was a problem hiding this comment.
Changed skills = list(skills) or None instead to avoid triggering deepagents skillsmidleware when the skills list is passed empty.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
mariadhakalUipath
force-pushed
the
feat/uipath-skills-for-advanced-agents
branch
from
5cee6e8 to
528a307
Compare
|
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
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.
Title: feat: add skills support to advanced agents with sandboxed workspace tools
Summary
Adds skills feature for advanced agents. Agent can act on it — run uip commands, review what it ran, and verify what it scaffolded.