AGENT SKILL

run-tests

Run the workspace test command for each detected stack and return pass or fail. Use when invoking /run-tests or when orchestrate-develop closes a story scope.

License 许可
MIT
最近更新
2026-11-08
Stars
★ 6
Forks
0

01 前言 Front Matter

name
run-tests
description
Run the workspace test command for each detected stack and return pass or fail. Use when invoking /run-tests or when orchestrate-develop closes a story scope.

02 资源文件 5

03 SKILL.md 正文

STOP - Read before ANY tool call

Read _shared/skill-boilerplate.md first. It holds the guardrails read, the session gate rule, the user sim rule, the one-step rule, the gate check, and the user chat language rule. The steps below are skill-specific only.

  1. This skill does not implement code and does not take a plan step When the parent is orchestrate-develop at scope close, the scope sim already authorizes the test command. Do not ask for a second sim only for that command.

Read-only modes (plan mode or ask): do not run a shell. Say that an edit mode is required.

Step -1b - Caveman Mode (Full cap)

When caveman is on, load {{TOOLKIT_ROOT}}/skills/_shared/caveman/CAVEMAN.md and apply the Full cap.


Skill: run-tests

Trigger

Negative signal: applies as defined in _shared/skill-boilerplate.md; the skill-specific signals are listed below.

/run-tests, or orchestrate-develop closing a story scope.

This skill does not replace test-coverage. It does not choose a stack in place of developer when the request is to implement.

Outcome

One report per detected stack, then an overall result. Tokens PASS, FAIL, and SKIPPED stay English. Sentences follow the user chat language.

Lazy-load

When Path
Stack order references/stack-detect.md
Command per stack references/commands.md
Report shape references/report.md
Must not references/must-not.md
Caveman {{TOOLKIT_ROOT}}/skills/_shared/caveman/CAVEMAN.md
Reference index reference.md

Do not load every developer guideline pack.

Never by default: do not preload every stack guideline or every references/*.md. Load one section for the detected stack.

Process

  1. Detect stacks with references/stack-detect.md. None: stop stack_not_detected and ask. Do not run a generic test.
  2. For each stack, run the command in references/commands.md. A missing command stops that stack with stack_test_command_missing.
  3. Emit references/report.md. Overall PASS only when every stack passed. A missing check is SKIPPED with a reason. It is not PASS.

04 同仓库更多 Skills