/skill-evaluation-workbench

name: skill-evaluation-workbench
description: Use when designing, running, debugging, or hardening deterministic eval suites for agent skills, prompts, tool workflows, or MCP-backed cases.

Skill Evaluation Workbench

When To Use

• A skill or prompt needs repeatable quality checks across models or configurations.
• A workflow needs file-based graders, command traces, or local artifact checks.
• A tool or MCP skill needs a hidden service fixture or sandboxed test workspace.
• A previous agent attempt failed and you need trace-driven diagnosis before editing instructions.

Requirements / Checks

• Confirm an eval runner exists locally before running anything. Do not install deps without approval.
• Prefer local deterministic graders over model-graded assertions.
• If Docker, remote models, API keys, or live services are involved, ask before execution.
• Treat traces, result files, preserved workspaces, and stdout as potentially sensitive.

Minimal Suite Structure

Every suite should have at least three cases:

| Case | Purpose | |---|---| | Positive (golden path) | Skill handles the normal use case correctly | | Edge case | Skill handles an important boundary condition | | Control (no-tool-needed) | Skill does not over-trigger on a clearly unrelated input |

Place fixtures in `cases/`, skill/reference material in `references/`, and grader scripts in `graders/`.

Grader Types

| Type | When to use | Deterministic? | |---|---|---| | File existence | Skill was supposed to create a file | Yes | | File content match | Output matches expected text or schema | Yes | | Command exit code | Script/tool succeeded | Yes | | JSON schema | Output is valid structured data | Yes | | Regex match | Output contains expected pattern | Yes | | Custom script | Complex logic not covered above | Yes (if written correctly) | | Model grader | Subjective quality judgment | No — use sparingly, pin model |

Workflow

1. **Define observable behavior** — state what the skill must produce: files, command args, JSON output, logs, or a safety refusal. If it's not observable, it can't be graded.

2. **Create the minimal suite** — one positive, one edge, one control case. Add more only after the minimal suite passes.

3. **Structure fixtures** — put case inputs in `cases/<case-name>/`, expected outputs in `cases/<case-name>/expected/`. For MCP-backed cases, use hidden service fixtures so server internals are not visible to the agent.

4. **Run smallest local suite first** — inspect the `result`, `summary`, trace, and workspace evidence from the first failure before editing anything.

5. **Classify the failure before editing**:

| Failure type | Fix target | |---|---| | Skill instructions are unclear | Edit `SKILL.md` | | Missing reference material | Add to `references/` | | Grader logic is brittle | Fix the grader script | | Fixture is unrealistic | Replace the test input | | Task is ambiguous | Clarify the task definition | | Actual product bug | File as a bug, not a skill fix |

6. **Edit only the classified cause**, then re-run the same case.

Safety Constraints

• Do not forward broad env vars into eval sandboxes — pass only named test variables.
• Do not print secrets in prompts, graders, traces, or artifacts.
• Do not mock a CLI or API unless the mock faithfully matches validation, output, and failure modes.
• Do not treat a green mock eval as proof that live integration works.
• Do not auto-enable host or user MCP configs inside eval containers.

Validation / Done Criteria

• Suite has deterministic pass/fail evidence for all cases.
• Failure triage points to a concrete cause before any edits are made.
• Re-run proves the targeted change fixed the failing case without breaking passing cases.
• Output includes: command used, model(s), trial count, failures, and artifact paths.

References

• `references/workbench-suite-model.md`