Skip to content

docs(uipath-agents): update eval skills for unified runtime eval commands - #2440

Open
AAgnihotry wants to merge 3 commits into
mainfrom
feat/eval-datasets-skills
Open

docs(uipath-agents): update eval skills for unified runtime eval commands#2440
AAgnihotry wants to merge 3 commits into
mainfrom
feat/eval-datasets-skills

Conversation

@AAgnihotry

Copy link
Copy Markdown
Contributor

Summary

  • Rewrites orchestrator-eval-run.md to cover all 30 uip or eval commands in one doc: execute-and-evaluate, evaluator CRUD, eval-set CRUD, evaluation CRUD, run results, and schedules
  • Replaces old run-offline-evals / --package-name / --is-low-code-agent references with execute-and-evaluate / --process-key / required --items+--evaluators
  • Adds typical workflow showing the full pipeline from creating evaluators through scheduling recurring runs
  • Updates evaluate.md nav and SKILL.md task table

Companion to UiPath/cli#3402.

Test plan

  • Verify doc links resolve correctly
  • Verify command examples match CLI implementation

🤖 Generated with Claude Code

…ands

Rewrites orchestrator-eval-run.md to cover the full `uip or eval` command
surface (30 commands across 6 groups):

- `execute-and-evaluate` replaces `run-offline-evals` — uses `--process-key`
  instead of `--package-name`/`--package-version`, requires `--items` and
  `--evaluators`, removes `--is-low-code-agent`
- New CRUD sections: evaluator, eval-set, evaluation (data points)
- Run results: list/get/results
- Schedules: create/list/get/update/pause/resume/delete
- Typical workflow showing the full create-evaluator → create-eval-set →
  add-data-points → run → check-results → schedule pipeline

Updates evaluate.md nav and SKILL.md task table.

Companion to UiPath/cli#3402.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
@github-actions

github-actions Bot commented Aug 4, 2026

Copy link
Copy Markdown
Contributor

Claude finished @AAgnihotry's task in 2m 50s —— View job


PR Review: docs(uipath-agents): update eval skills for unified runtime eval commands

  • Read review criteria and project conventions
  • Read full PR diff
  • Read changed files in full context
  • Check links, cross-skill refs, secrets
  • Post review findings

Summary

Rewrites orchestrator-eval-run.md from a single-command doc (run-offline-evals) into a comprehensive reference covering all 30 uip or eval subcommands: execute-and-evaluate, evaluator/eval-set/evaluation CRUD, run results, and schedules. Updates the SKILL.md task table and evaluate.md nav link to match. All old --package-name/--is-low-code-agent/run-offline-evals references are cleanly removed.

Change-by-Change Review

1. skills/uipath-agents/SKILL.md (line 70)

Severity: OK

Task table row updated from "Run offline evals for a published Orchestrator package" to "Manage runtime evals for a published Orchestrator package". Description column now lists the key capabilities (execute-and-evaluate, CRUD, run results, schedules). Clean, accurate.

2. skills/uipath-agents/references/lowcode/evaluations/evaluate.md (line 34)

Severity: OK

Nav link updated from "Orchestrator Package Offline Evals" to "Orchestrator Runtime Eval Commands" with an expanded description. Correctly reflects the new scope of the linked doc.

3. skills/uipath-agents/references/lowcode/evaluations/orchestrator-eval-run.md

Severity: Low — two minor observations, otherwise well-structured.

  • Typical Workflow steps 3 and 4 are semantically redundant (lines 353–365): Step 3 creates a data point via evaluation create with inputs: {"input":"hello"}, but step 4 passes the same data inline via --items to execute-and-evaluate rather than referencing the eval set created in step 2 via --eval-set-id. An agent following this workflow would create the data point via CRUD and then re-specify it inline, which is confusing. Consider either:

    • (a) having step 4 use --eval-set-id "$EVAL_SET_ID" instead of --items/--evaluators to show the CRUD-first path, or
    • (b) dropping step 3 and showing the inline-only path.

    Showing both paths in one pipeline conflates two distinct workflows — CRUD-then-reference vs. inline-only.

    Fix this →

  • Missing Anti-patterns section (end of file): The old version had an Anti-patterns section covering common mistakes. The rewrite drops it entirely. Per the skill structure rules, anti-patterns sections prevent expensive agent mistakes and should be preserved when relevant. For this doc, useful anti-patterns would include:

    • Don't pass --items/--evaluators inline when an eval set already exists (duplicates data, risks drift).
    • Don't forget --folder-key on create commands when not using the personal workspace.
    • Don't mix CRUD-created evaluators with inline evaluator configs that have empty evaluatorConfig objects (step 4 shows "evaluatorConfig":{} which may not carry the prompt/model).

    Fix this →

What's Missing

  • Anti-patterns section in orchestrator-eval-run.md — the old version had one; the rewrite drops it. This is a quality gap for agent guidance (Medium).
  • No test tasks cover the uip or eval Orchestrator commands. No existing tasks reference run-offline-evals or execute-and-evaluate. This is pre-existing — the PR doesn't regress coverage, but the substantial expansion of documented commands (from 1 to ~30) widens the gap (Medium, not blocking).

Area Ratings

Area Status Notes
Frontmatter OK No frontmatter changes beyond the task table
E2E Tests Issue No test tasks cover orchestrator eval commands (pre-existing gap, not introduced by this PR)
Skill Body OK Task table row accurately reflects new doc scope
References & Assets OK All internal links resolve; nav updated consistently; no stale run-offline-evals refs remain anywhere in skills/uipath-agents/
Repo Hygiene OK No secrets, no cross-skill dependencies, no personal paths, changes scoped to one skill

Issues for Manual Review

  • CLI command accuracy: The 30 commands documented (evaluator/eval-set/evaluation/schedule CRUD, execute-and-evaluate, run results) should be verified against the actual CLI implementation in UiPath/cli#3402. Flag names, required vs. optional status, and output codes are taken at face value in this review.
  • Typical Workflow step 4: The evaluatorConfig: {} empty object in the inline --evaluators JSON may not be valid at runtime — verify whether the API accepts an empty config or requires the full prompt/model/type fields.

Conclusion

Clean, well-structured rewrite that properly replaces the old single-command doc with comprehensive CRUD coverage. The two Low-severity items (workflow step redundancy and missing anti-patterns) are worth addressing but not blocking. No Critical or High issues found. Approve with minor suggestions.


AAgnihotry and others added 2 commits August 3, 2026 23:39
Adds --limit/--offset options and Pagination field to evaluator list,
eval-set list, evaluation list, and run list command docs to match CLI.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Fix workflow step redundancy: step 4 now references the eval set
  created in step 2 via --eval-set-id instead of duplicating data inline
- Fix empty evaluatorConfig: {} in workflow example — now shows the full
  config with prompt, model, and targetOutputKey
- Add Anti-patterns section covering empty configs, same-as-agent, missing
  folder-key, CRUD vs inline mixing, and cross-process evaluator reuse

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Copilot AI left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Pull request overview

Updates the uipath-agents documentation to reflect the unified uip or eval runtime evaluation command surface for agents deployed as Orchestrator packages, replacing the older “offline eval” guidance and updating internal navigation to the new reference.

Changes:

  • Rewrote orchestrator-eval-run.md into a consolidated reference covering uip or eval command groups (execute-and-evaluate, CRUD, results, schedules) and a typical end-to-end workflow.
  • Updated the skill’s task table entry to point to the new runtime-eval command set and scope.
  • Updated the evaluations navigation to reference the renamed/repurposed Orchestrator runtime eval doc.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 3 comments.

File Description
skills/uipath-agents/SKILL.md Updates the task table entry from “offline evals” to runtime eval management coverage.
skills/uipath-agents/references/lowcode/evaluations/orchestrator-eval-run.md Replaces the prior offline-eval doc with a unified runtime-eval command reference, examples, and troubleshooting.
skills/uipath-agents/references/lowcode/evaluations/evaluate.md Updates nav text to the new Orchestrator runtime eval documentation entry.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

# Orchestrator Runtime Eval Commands

Submit offline evaluation runs for low-code agents published as Orchestrator packages.
Manage and run runtime evaluations for agents published as Orchestrator packages. All commands are scoped by `--process-key` (deployment GUID) and hit the agents runtime API.
Comment on lines +238 to +252
[--evaluation-criterias <json>] \
[--tenant <tenant>] \
--output json
```

| Flag | Required | Description |
|------|----------|-------------|
| `--process-key` | Yes | Process key (GUID) |
| `--eval-set-id` | Yes | Eval set ID (GUID) |
| `--folder-key` | Yes | Folder key (GUID) |
| `--name` | Yes | Data point name |
| `--inputs` | Yes | Input values as JSON |
| `--expected-output` | No | Expected output as JSON (for output evaluators) |
| `--expected-behavior` | No | Expected agent behavior (for trajectory evaluators) |
| `--evaluation-criterias` | No | Per-evaluator criteria overrides as JSON |
Comment on lines +364 to +370
# 4. Run the eval referencing the eval set
uip or eval execute-and-evaluate \
--process-key "$PROCESS_KEY" \
--eval-set-id "$EVAL_SET_ID" \
--items '[{"id":"i1","name":"Greeting test","inputs":{"input":"hello"},"expectedOutput":{"content":"Hi there!"},"expectedBehavior":""}]' \
--evaluators '[{"id":"'"$EVALUATOR_ID"'","version":"","evaluatorTypeId":"5","evaluatorConfig":{"name":"Semantic","type":5,"category":1,"prompt":"Score 0-100...","model":"gpt-4.1-2025-04-14","targetOutputKey":"*"}}]' \
--output json
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants