task-execution

name	task-execution
description	Execute general user tasks with tool-backed evidence and reusable skill feedback
version	1.0.0
tags	["task-execution","verification","skill-iteration"]
triggers	["execute this task","investigate and fix","verify with tools","diagnose the problem","turn this workflow into a skill","执行任务","排查问题","验证结果"]
update_section	## Reference Data
execution_contract	{"version":1,"min_pass_score":0.7,"rewrite_below_score":0.55,"disable_below_score":0.25,"min_samples_for_rewrite":3,"min_samples_for_disable":2,"criteria":[{"name":"goal_completion","description":"The response completes the user's requested operational outcome.","weight":0.4},{"name":"tool_grounding","description":"Claims and decisions are grounded in available tool output or explicit limits.","weight":0.35},{"name":"durable_learning","description":"Reusable lessons are captured or routed to skill creation when appropriate.","weight":0.25}],"observable_signals":[{"name":"relevant_tool_trace","description":"Relevant tool calls or an explicit no-tool limitation are present.","source":"tool_trace","weight":0.3}],"failure_modes":["skipped_required_verification","hallucinated_external_state","failed_to_capture_reusable_workflow"]}
outcome_contract	{"entity":"task","primary_kpi":"success_score","supporting_kpis":["reuse_count"],"dimensions":["channel","outcome"],"window":{"current_days":7,"baseline_days":7},"min_sample_size":2,"rewrite_policy":{"target":90,"min_delta":5,"require_segment":true,"severe_regression_delta":15},"rollback_policy":{"max_negative_delta":10},"reference_update_policy":{"max_segments":3,"max_exemplars":2}}

Instructions

Use this skill for general operational tasks:

1. Identify the requested outcome and constraints.
2. Use available tools for evidence: logs, files, HTTP calls, shell commands, tests, or database queries.
3. Keep the response grounded in what was actually checked or changed.
4. If the task reveals a repeatable capability, create, update, or reuse a skill package.
5. Preserve useful lessons in references so future runs improve.

Reference Data

This section is replaced by evidune run after each iteration cycle.