Is docling-ingest safe to install?

SaferSkills independently audited docling-ingest (Agent Skill) and scored it 100/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 0 high-severity and 0 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.

CategoryWeightCategory scoreContribution

Securityprompt, exec, net, exfil, eval

35%

100

35.0 pts

Supply chainhash, typosquat, maintainer, lockfile

20%

100

20.0 pts

Maintenancestaleness, pinning, CI

15%

100

15.0 pts

TransparencySKILL.md, perms, README

15%

100

15.0 pts

Communityinstalls, verify, response

15%

100

15.0 pts

Findings & checks · 0 flagged

Securityscore 100 · 0 findings

✓—All security checks passedNo findings in this category for the latest scan.pass

Supply chainscore 100 · 0 findings

✓—All supply chain checks passedNo findings in this category for the latest scan.pass

Maintenancescore 100 · 0 findings

✓—All maintenance checks passedNo findings in this category for the latest scan.pass

Transparencyscore 100 · 0 findings

✓—All transparency checks passedNo findings in this category for the latest scan.pass

Communityscore 100 · 0 findings

✓—All community checks passedNo findings in this category for the latest scan.pass

The primary manifest — the file an agent reads to learn what this artifact does.

skills/docling-ingest/SKILL.md · 1 file

skills/docling-ingest/SKILL.md1.7 KB · Markdown

name: docling-ingest

description: >-

用 Docling 将 PDF/DOCX/PPTX/XLSX 等转为 Markdown，写入 knowledge-base/raw/，

供后续 wiki ingest。口令：「docling 转一下」「PDF 进知识库」。需本机已 pip install docling。

disable-model-invocation: false

Docling 文档解析

项目：docling-project/docling — Get your documents ready for gen AI.

分工

步骤	Skill / 动作
二进制 → Markdown	本 Skill
Markdown → 互链 wiki	按你的 wiki schema 做 ingest

前置条件

Python ≥ 3.10
安装：pip install docling（或 uv pip install docling）
敏感文档：本地执行，不上传第三方解析服务。

用法

在 仓库根目录（含 knowledge-base/raw/）：

# 单文件 → knowledge-base/raw/<basename>.md
bash skills/docling-ingest/scripts/docling-to-raw.sh "/path/to/file.pdf"

# 或 Python 直接调
python3 skills/docling-ingest/scripts/docling_convert.py --file "/path/to/file.pdf"

成功后提示用户：「是否做 wiki ingest？」

Agent 规则

必须先有明确文件路径；不得猜测仓库内任意 PDF。
输出写入 `knowledge-base/raw/`，文件名建议 YYYY-MM-DD-<主题>.md。
不假装「已精读」——转换后仍须 ingest 或人工扫一眼表格/数字。
含 客户隐私、未脱敏合同 → 仅本地 raw，禁止上传公网。

与「不擅长」的边界

Docling 解决 可读文本与结构；扫描极差、手写、纯图无 OCR 的页 → 标注 待人工 或 低置信度，不编造正文。

docling-ingest— agent skill

Is docling-ingest safe to install?

Docling 文档解析

分工

前置条件

用法

Agent 规则

与「不擅长」的边界

Audit the pieces. Scan the whole. Decide.