一键在 Manus 中运行任何 Skill

bash-tools

星标0

分支0

更新时间2026年6月22日 06:34

Lists the agent bash utility scripts in scripts/agent/ with their headers and usage lines. Use this at the start of a session, when decomposing a new task, or whenever you catch yourself about to reach for raw tool calls that one of these scripts already composes. Refreshes your working memory of the bash layer so you don't reinvent what's already there.

安装

用 Codex 或 Claude 帮你安装复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。

在 Manus 中运行

来源

JStepUX

JStepUX/aix-boilerplate

打开 GitHub 仓库查看创作者相关仓库

下载

在 Manus 中运行

Agent Bash Tools — Live Index

The scripts/agent/ folder contains bash utilities that collapse common multi-tool-call patterns into single invocations. They produce deterministic output and are the preferred entry point for discovery, verification, and context-gathering tasks. Reaching for raw Grep/Read/Glob when one of these already exists wastes context and misses the stable format.

Step 1: List the current tools

Run this exact command and display the output:

for f in scripts/agent/*.sh; do
  name=$(basename "$f")
  [[ "$name" == "_common.sh" ]] && continue
  desc=$(sed -n '2p' "$f" | sed -E "s/^# *//; s/^${name} *[^A-Za-z]*//")
  usage=$(sed -n '3p' "$f" | sed -E 's/^# *//')
  printf '%s\n  %s\n  %s\n\n' "$name" "$desc" "$usage"
done

This reads line 2 (description) and line 3 (usage) of each script, which is the canonical header format in this folder. If a script is missing that header, the output will be blank for that script — flag it so the header can be added.

Step 2: Match to the current task

After listing, identify in one or two sentences which scripts apply to the task you're about to do, and reach for them first. Examples:

"I'm about to understand an unfamiliar file" → file-context.sh
"I'm about to commit or open a PR" → git-context.sh then health-check.sh
"I'm about to hunt for a feature across the codebase" → related-files.sh
"I'm about to figure out what depends on a module" → trace-imports.sh

If none apply, proceed with raw tool calls and note why none fit — that gap is data for whether a new script earns its spot later.

Step 3: Note anything surprising

If a script has appeared or disappeared since you last saw the list, mention it to the user. The folder is the source of truth, not anyone's memory of it.

同仓库更多 Skills

同仓库

pre-commit-qa

JStepUX/aix-boilerplate

A mandatory quality gate that enforces test coverage, localization, documentation, spec archival, and commit hygiene after any implementation work. Use this skill whenever code changes are complete and about to be committed — trigger on phrases like "I'm done", "ready to commit", "finished implementing", "task complete", "wrap up", "ship it", "PR ready", or any signal that implementation is finished and the work is moving toward version control. Also trigger when reviewing someone else's completed work before merge. This skill should fire even if the user doesn't explicitly ask for a quality check — if implementation just finished, this gate applies. Do NOT skip steps. Do NOT self-certify. Every item requires evidence.

2026-06-220

recon

JStepUX/aix-boilerplate

Discovery-phase skill for symptom-shaped tasks. Use when the user hands you a vague problem, a bug report, a console error, or a "this is broken" style request rather than a well-specified feature. Produces a grounded problem brief you can either act on directly or hand to the project-coordinator agent as a sharp, pre-diagnosed task spec. Closes the gap between "here's a symptom" and "here's a decomposable plan".

2026-06-220

release-prep

JStepUX/aix-boilerplate

Prepares a release. Reads the changelog entries that pre-commit-qa has accumulated since the last release marker, cross-checks against git for any bypassed commits, proposes a semver bump, and — on explicit user approval — inserts a release marker, bumps the version, commits, and tags. Does not push and does not deploy. Use before running the project's deploy command, or when the user signals release intent ("ready to deploy", "time to ship", "cut a release", "write the changelog", "release prep"). The authoring side of the release pipeline — complements `pre-commit-qa` (which gates individual commits) by gating the batch that becomes a release.

2026-06-220

name	bash-tools
description	Lists the agent bash utility scripts in scripts/agent/ with their headers and usage lines. Use this at the start of a session, when decomposing a new task, or whenever you catch yourself about to reach for raw tool calls that one of these scripts already composes. Refreshes your working memory of the bash layer so you don't reinvent what's already there.

Agent Bash Tools — Live Index

Step 1: List the current tools

Run this exact command and display the output:

for f in scripts/agent/*.sh; do
  name=$(basename "$f")
  [[ "$name" == "_common.sh" ]] && continue
  desc=$(sed -n '2p' "$f" | sed -E "s/^# *//; s/^${name} *[^A-Za-z]*//")
  usage=$(sed -n '3p' "$f" | sed -E 's/^# *//')
  printf '%s\n  %s\n  %s\n\n' "$name" "$desc" "$usage"
done

Step 2: Match to the current task

After listing, identify in one or two sentences which scripts apply to the task you're about to do, and reach for them first. Examples:

"I'm about to understand an unfamiliar file" → file-context.sh
"I'm about to commit or open a PR" → git-context.sh then health-check.sh
"I'm about to hunt for a feature across the codebase" → related-files.sh
"I'm about to figure out what depends on a module" → trace-imports.sh

If none apply, proceed with raw tool calls and note why none fit — that gap is data for whether a new script earns its spot later.

Step 3: Note anything surprising

If a script has appeared or disappeared since you last saw the list, mention it to the user. The folder is the source of truth, not anyone's memory of it.