ecc-eval-harness

name	ecc-eval-harness
description	Domain skill for 'eval-harness'. Use when requests involve evaluation, benchmarking, quality scoring, and regression checks; agent harness configuration, performance, and reliability tuning.

ecc-eval-harness

Apply the skill as a concrete workflow rather than static guidance.

Workflow

1. Clarify the task objective and constraints.
2. Load references/upstream-summary.md for source intent.
3. Translate upstream harness-specific semantics into OpenClaw tool steps.
4. Execute in smallest safe increments.
5. Verify outcomes and report evidence.

Output Requirements

• Objective
• Actions taken
• Verification performed
• Risks/limitations
• Next recommended step

Rules

• Prefer deterministic scripts/commands when possible.
• Avoid unverified claims; provide evidence.
• If uncertain, state assumptions explicitly.

ECC Parity Notes

• This skill is adapted from ECC intent and translated to OpenClaw-native tool workflows.
• Prefer reproducible, tool-backed execution over harness-specific slash semantics.
• Keep outputs evidence-based and verification-first.

OpenClaw Constraints

• If upstream flow assumes unavailable hooks/tools, use documented fallback scripts in openclaw-ecc/scripts/.
• Mark non-applicable checks as SKIP rather than hard-failing when environment support is absent.