docs(uipath-agents): document eval set schema key and filename convention (UV-14439)

[mjnovice]

Why

UV-14439 — session retrospective reported doc gaps in low-code eval analysis. Most items in the ticket are already addressed by the lowcode/evaluations/ reference tree that landed since the ticket was filed:

• uip agent eval run start/status/results/list/compare with --set (required) and --verbose → running-evaluations.md
• SKILL.md task-navigation rows for low-code evals + Orchestrator offline evals
• Cloud-only run results → evaluate.md Prerequisites
• Pre-flight tenant check → authentication.md

What remains (this PR)

Two schema notes in evaluation-sets.md the retrospective flagged as wasted-introspection causes:

• Test-case array is evaluations[], not testCases[]
• Studio Web filename convention evaluation-set-<unix-ms>.json; match sets by name/id, not filename pattern

🤖 Generated with Claude Code

[github-actions]

Claude finished @mjnovice's task in 54s —— View job

---

PR Review: docs(uipath-agents) — eval set schema notes

• Read review criteria and project conventions
• Read full diff and changed file in context
• Check links, cross-skill references, secrets
• Post review

---

Summary

Adds two schema notes to evaluation-sets.md clarifying that the test-case array key is evaluations[] (not testCases[]) and that Studio Web uses timestamp-based filenames for eval sets. Both address documented wasted-introspection causes from a session retrospective (UV-14439).

Change-by-Change Review

1. skills/uipath-agents/references/lowcode/evaluations/evaluation-sets.md (+5/−0)

Severity: OK

Two bullet points inserted at line 146–149, between the JSON example block and the existing source field paragraph. Placement is correct — the notes annotate the JSON schema shown directly above.

Content accuracy:

• The evaluations[] key matches the JSON example at line 120. Good that it calls out the likely false guess (testCases[]).
• The filename convention note (evaluation-set-<unix-ms>.json) is actionable — tells the agent to match by name/id rather than filename pattern.
• The inline CLI command (uip agent eval list --set ... --output json) is consistent with the same command documented at line 66 and follows the --output json convention.

Tone and compression comply with token-optimization rules — prescriptive, no filler, fragments where appropriate.

No cross-skill references. No secrets or personal paths. No broken links.

What's Missing

Nothing identified.

Area Ratings

Area	Status	Notes
Frontmatter	OK	Not touched; no changes needed
E2E Tests	OK	Doc-only change to existing skill; no new capabilities requiring test coverage
Skill Body	OK	Not touched
References & Assets	OK	Two well-placed schema notes, consistent with surrounding content
Repo Hygiene	OK	CODEOWNERS already covers skills/uipath-agents/; scoped change, no secrets

Issues for Manual Review

• The evaluation-set-<unix-ms>.json filename convention is domain knowledge from Studio Web behavior — reviewer should confirm this is the current/stable naming pattern and not subject to change.

Conclusion

Clean, well-scoped documentation addition. No issues found. Approve.

---