Ghost-Scan-Code：自动化代码安全扫描工具，快速发现仓库漏洞

ghost-scan-code by ghostsecurity/skills

1,000 周安装量

370 GitHub Stars

GitHub

安装命令

npx skills add https://github.com/ghostsecurity/skills --skill ghost-scan-code

自动化开发运维安全

🇨🇳中文介绍

查找问题

您可以在代码仓库中发现安全问题。此技能规划要扫描的漏洞向量，然后对每个项目执行这些扫描。

输入

depth : quick（默认）、balanced 或 full — 可通过 $ARGUMENTS 覆盖

$ARGUMENTS

注意： 如果提供了参数，可用于自定义扫描工作流程。例如，如果用户指定了特定的向量集、向量数量、特定的候选文件、需要重点关注的区域、候选文件数量等，请确保将相关细节传递给技能中的相应步骤。

支持文件

循环脚本: scripts/loop.sh
扫描标准: criteria/index.yaml

步骤 1：设置

计算仓库特定的输出目录：

repo_name=$(basename "$(pwd)") && remote_url=$(git remote get-url origin 2>/dev/null || pwd) && short_hash=$(printf '%s' "$remote_url" | git hash-object --stdin | cut -c1-8) && repo_id="${repo_name}-${short_hash}" && short_sha=$(git rev-parse --short HEAD 2>/dev/null || date +%Y%m%d) && ghost_repo_dir="$HOME/.ghost/repos/${repo_id}" && scan_dir="${ghost_repo_dir}/scans/${short_sha}/code" && cache_dir="${ghost_repo_dir}/cache" && mkdir -p "$scan_dir" && echo "scan_dir=$scan_dir cache_dir=$cache_dir"

广告位招租

在这里展示您的产品或服务

触达数万 AI 开发者，精准高效

联系我们

步骤 2：规划扫描

如果 $scan_dir/plan.md 已存在，则跳至下一步。

否则，使用 scripts/loop.sh 运行规划器：

bash <path-to-loop.sh> $scan_dir planner.md "- depth: <depth>
- arguments: <relevant argument overrides if any, otherwise omit>" 1 $cache_dir

使用 10 分钟超时。如果命令超时，请重新运行 — 脚本将从上次中断处恢复。如果连续 3 次因相同错误而失败，则停止并报告失败。

验证： 在继续之前，确保 $scan_dir/plan.md 存在且至少包含一个 ## Project: 部分。

步骤 3：提名文件

如果 $scan_dir/nominations.md 不存在，则通过读取 $scan_dir/plan.md 来生成它，并为每个项目部分（## Project: <base_path> (<type>)）解析“推荐扫描”表格。对于每一行，提取“代理”和“向量”列。写入 $scan_dir/nominations.md — 每行对应一个（项目、代理、向量）组合。跳过扫描表格为空的项目。

# 提名

- [ ] <base_path> (<type>) | <agent> | <vector>
- [ ] <base_path> (<type>) | <agent> | <vector>
...

如果 $scan_dir/nominations.md 已存在，则将每个顶级任务 - [x] 更改为 - [ ]。保持每个项目下所有缩进行/子任务不变。

使用 scripts/loop.sh：

bash <path-to-loop.sh> $scan_dir nominator.md "- depth: <depth>
- arguments: <relevant argument overrides if any, otherwise omit>" 5 $cache_dir

使用 10 分钟超时。如果命令超时，请重新运行 — 脚本将从上次中断处恢复。如果连续 3 次因相同错误而失败，则停止并报告失败。

验证： 在继续之前，确保 $scan_dir/nominations.md 至少包含一个 - [x] 行。

步骤 4：分析提名文件

读取 $scan_dir/nominations.md。对于每个在已勾选的 - [x] 行下的候选文件，将其追加到 $scan_dir/analyses.md（跳过已列在 analyses.md 中的候选文件）。

- [ ] <base_path> (<type>) | <agent> | <vector> | <candidate_file>

创建发现目录：

mkdir -p $scan_dir/findings

使用 scripts/loop.sh：

bash <path-to-loop.sh> $scan_dir analyzer.md "" 5 $cache_dir

使用 10 分钟超时。如果命令超时，请重新运行 — 脚本将从上次中断处恢复。如果连续 3 次因相同错误而失败，则停止并报告失败。

验证： 在继续之前，确保 $scan_dir/analyses.md 至少包含一个 - [x] 行。

步骤 5：验证发现

列出 $scan_dir/findings/ 中的所有 .md 文件。如果不存在任何文件，则写入 no-findings.md 摘要并停止。

使用 scripts/loop.sh：

bash <path-to-loop.sh> $scan_dir verifier.md "" 5 $cache_dir

使用 10 分钟超时。如果命令超时，请重新运行 — 脚本将从上次中断处恢复。如果连续 3 次因相同错误而失败，则停止并报告失败。

所有步骤完成后，报告扫描结果：

列出 $scan_dir/findings/ 中的所有发现文件
统计已验证与已拒绝的发现数量
向用户呈现摘要

🇺🇸English

Find Issues

You find security issues in a repository. This skill plans which vulnerability vectors to scan, then executes those scans against each project.

Inputs

depth : quick (default), balanced, or full — override via $ARGUMENTS

$ARGUMENTS

Note: Arguments passed can be used to customize the scan workflow if provided. For example, if the user specifies a specific set of vectors, count of vectors, specific candidate files, areas to focus on, count of candidate files, etc., ensure the relevant details are passed to the relevant steps in the skill.

Supporting files

Loop script: scripts/loop.sh
Scan criteria: criteria/index.yaml

Step 1: Setup

Compute the repo-specific output directory:

repo_name=$(basename "$(pwd)") && remote_url=$(git remote get-url origin 2>/dev/null || pwd) && short_hash=$(printf '%s' "$remote_url" | git hash-object --stdin | cut -c1-8) && repo_id="${repo_name}-${short_hash}" && short_sha=$(git rev-parse --short HEAD 2>/dev/null || date +%Y%m%d) && ghost_repo_dir="$HOME/.ghost/repos/${repo_id}" && scan_dir="${ghost_repo_dir}/scans/${short_sha}/code" && cache_dir="${ghost_repo_dir}/cache" && mkdir -p "$scan_dir" && echo "scan_dir=$scan_dir cache_dir=$cache_dir"

Read $cache_dir/repo.md — if missing, run the repo-context skill first and then continue.
Read criteria/index.yaml to get the valid agent→vector mappings per project type
Set depth to quick if not provided
If depth is full, warn the user that a full scan uses significantly more tokens and ask them to confirm before proceeding. If they decline, fall back to balanced.

Step 2: Plan Scans

If $scan_dir/plan.md already exists, skip to the next step.

Otherwise, run the planner using scripts/loop.sh:

bash <path-to-loop.sh> $scan_dir planner.md "- depth: <depth>
- arguments: <relevant argument overrides if any, otherwise omit>" 1 $cache_dir

Use a 10-minute timeout. If the command times out, re-run it — the script resumes from where it left off. If it fails 3 times consecutively with the same error, stop and report the failure.

Verify: $scan_dir/plan.md exists and contains at least one ## Project: section before proceeding.

Step 3: Nominate Files

If $scan_dir/nominations.md does not exist, generate it by reading $scan_dir/plan.md and for each project section (## Project: <base_path> (<type>)), parse the Recommended Scans table. For each row, extract the Agent and Vector columns. Write $scan_dir/nominations.md - one line per (project, agent, vector) combination. Skip projects with empty scan tables.

# Nominations

- [ ] <base_path> (<type>) | <agent> | <vector>
- [ ] <base_path> (<type>) | <agent> | <vector>
...

If $scan_dir/nominations.md already exists, change every top level task - [x] to - [ ]. Keep all indented lines/subtasks beneath each item unchanged.

Run nomination script

Using scripts/loop.sh:

bash <path-to-loop.sh> $scan_dir nominator.md "- depth: <depth>
- arguments: <relevant argument overrides if any, otherwise omit>" 5 $cache_dir

Use a 10-minute timeout. If the command times out, re-run it — the script resumes from where it left off. If it fails 3 times consecutively with the same error, stop and report the failure.

Verify: $scan_dir/nominations.md contains at least one - [x] line before proceeding.

Step 4: Analyze Nominated Files

Read $scan_dir/nominations.md. For each candidate file under a checked - [x] line, append to $scan_dir/analyses.md (skip candidates already listed in analyses.md).

- [ ] <base_path> (<type>) | <agent> | <vector> | <candidate_file>

Create the findings directory:

mkdir -p $scan_dir/findings

Run analysis script

Using scripts/loop.sh:

bash <path-to-loop.sh> $scan_dir analyzer.md "" 5 $cache_dir

Use a 10-minute timeout. If the command times out, re-run it — the script resumes from where it left off. If it fails 3 times consecutively with the same error, stop and report the failure.

Verify: $scan_dir/analyses.md contains at least one - [x] line before proceeding.

Step 5: Verify Findings

List all .md files in $scan_dir/findings/. If none exist, write a no-findings.md summary and stop.

Using scripts/loop.sh:

bash <path-to-loop.sh> $scan_dir verifier.md "" 5 $cache_dir

Use a 10-minute timeout. If the command times out, re-run it — the script resumes from where it left off. If it fails 3 times consecutively with the same error, stop and report the failure.

Completion

After all steps complete, report the scan results:

List all finding files in $scan_dir/findings/
Count verified vs rejected findings
Present a summary to the user

Weekly Installs

1.0K

Repository

ghostsecurity/skills

GitHub Stars

368

First Seen

Feb 20, 2026

Security Audits

Gen Agent Trust HubPass SocketPass SnykPass

Installed on

claude-code756

codex403

gemini-cli402

opencode402

github-copilot401

kimi-cli401

Ghost-Scan-Code：自动化代码安全扫描工具，快速发现仓库漏洞

🇨🇳中文介绍

查找问题

输入

支持文件

步骤 1：设置

相关 Skills

步骤 2：规划扫描

步骤 3：提名文件

运行提名脚本

步骤 4：分析提名文件

运行分析脚本

步骤 5：验证发现

完成

🇺🇸English

Find Issues

Inputs

Supporting files

Step 1: Setup

Step 2: Plan Scans

Step 3: Nominate Files

Run nomination script

Step 4: Analyze Nominated Files

Run analysis script

Step 5: Verify Findings

Completion

最新 Skills