Convert Office documents and PDFs (incl. scanned, via CPU-first routing OCR with lightweight models: PP-DocLayout-L layout, RapidOCR text, SLANet tables, FormulaNet formulas) to structurally-formatted Markdown. CLI + dsh agent tool (md_convert).
Install verification: runtime verified (L5) · tested on dsh 0.2.1-alpha.2 · last checked Install verification details →
What it doesAI
Converts Office docs and PDFs (incl. scanned) into structured Markdown via lightweight OCR models, with CLI and agent tool.
- CPU-first OCR routing with lightweight models
- Detects layout, text, tables, and formulas
- CLI and dsh agent tool (md_convert)
Capability tags are AI-inferred or rule-extracted from the README (see badge) — vocabulary-limited, not install-verified. AI = inferred; README = literal keyword from the readme.
Installation
dsh plugin --profile web add github:yakoylp/dsh-md-convert#v0.5.6Install method: GitHub · runtime verified (L5) · verified on dsh 0.2.1-alpha.2
Verification & Compatibility
Grouped by plugin version: each card lists the dsh versions actually tested against that artifact version and their install verdicts (L5 wins). Untested combinations are omitted — absence means untested, never assumed compatible. L1–L3 are a plugin-level static archive (independent of the dsh version — the same verdict on every dsh row): they sit in each row's level grid next to L4/L5, marked "static", while their evidence is stored once per plugin inside that row's "checks & evidence" instead of being duplicated per version.
External credentials (plugin-level): none detected
○ The author declares no dsh version requirement — install is not blocked by version.
| plugin ＼ dsh | dsh 0.2.1-alpha.2 | dsh 0.1.7-rc.1 |
|---|---|---|
| v0.5.6 | ✓ | · |
v0.5.6github✓ 1 runtime-verified2026-10-18
https://github.com/yakoylp/dsh-md-convert
✓dsh 0.2.1-alpha.22026-10-10L5 · runtime verified
- • Node.js: v24.14.0
- • DSH: 0.2.1-alpha.2
- • Plugin artifact: github · https://github.com/yakoylp/dsh-md-convert#v0.5.6
ID: dsh-md-convert@v0.5.6@dsh0.2.1-alpha.2 · profile: l4-sandbox
Issued: 2026-10-10 · expires: 2026-10-17
View checks & evidence
active in loader tree (150 entries)
L1–L3 are plugin-level static conclusions (baseline, identical across dsh versions): evidence is stored once per plugin.
Only the latest 3 published versions are shown; source-snapshot (commit hash) groups, unstated groups and older versions are omitted — the full archive stays in the data layer (snapshot/unstated groups appear only when a plugin has no published version).
Author badges
Embed the official badges in your README to showcase security & install-test status:
`[](https://www.dsh.so/artifact/dsh-md-convert/)``[](https://www.dsh.so/artifact/dsh-md-convert/)`Security Report
Two automated layers, neither a manual review: a static scan (per-plugin rule checks) and a model-driven dynamic security audit that covers only the subset static-flagged at warn or above and never executes plugin code.
Static low, dynamic medium — stricter wins → overall medium.
v0.5.6dsh manifestv0.5.6git2026-10-21v0.5.6same versionclean · gitVet is an AST static scan (npm artifact or git source).
show static findings (8)
- warningchildProcess · code-exec
lib/core/deps.js:47const r = spawnSync(py, ["--version"], { encoding: "utf8", timeout: 15_000, windowsHide: true }); - warningchildProcess · code-exec
lib/core/deps.js:63const r = spawnSync(python, ["-c", probe], { encoding: "utf8", timeout: 60_000, windowsHide: true }); - warningchildProcess · code-exec
lib/core/deps.js:180const r = spawnSync(python, ["-c", probe], { encoding: "utf8", timeout: 60_000, windowsHide: true }); - warningchildProcess · code-exec
lib/core/legacy.js:102const r = spawnSync("powershell.exe", ["-NoProfile", "-NonInteractive", "-ExecutionPolicy", "Bypass", "-File", scriptPath, ...args], { - warningchildProcess · code-exec
lib/core/legacy.js:145const r = spawnSync("powershell.exe", ["-NoProfile", "-NonInteractive", "-ExecutionPolicy", "Bypass", "-Command", script], { - warningchildProcess · code-exec
lib/core/legacy.js:176const r = spawnSync(cmd, ["--version"], { encoding: "utf8", timeout: 30_000, windowsHide: true }); - warningshellExec · shell
lib/core/legacy.js:124try { execFileSync("powershell.exe", ["-NoProfile", "-Command", "Start-Sleep -Milliseconds 2500"], { windowsHide: true, timeout: 10_000 }); - warningchildProcess · code-exec
lib/core/spawn.js:29child = spawn(cmd, args, { windowsHide, env });
show 2 audit findings
- mediummd_convert silently pip-installs OCR packages into the host Python, with --break-system-packages fallback
lib/core/deps.js:88The agent tool path enables this by default (lib/index.js: `autoInstallDeps: config.autoInstallDeps !== false`), so a single md_convert call on a scanned PDF runs pip against the host interpreter without any user confirmation. On Debian/Ubuntu system Python the error path retries with `--break-system-packages` (deps.js:97), deliberately overriding the OS package-manager protection, and installs the unpinned latest versions of paddlepaddle/paddleocr/paddlex[ocr]/pypdfium2/rapidocr/onnxruntime. The tool description shown to the model does not disclose the installation, and the probe failing (non
- lowmd_convert writes output to any caller-specified directory
lib/index.js:96outDir comes straight from the tool arguments and is resolved with path.resolve, so an absolute path bypasses the workspace (the doc comment claims output stays 'in the workspace'). With overwrite defaulting to true (convert.js DEFAULT_OPTS.overwrite = true) the derived `<basename>.md` file can overwrite a pre-existing Markdown file anywhere the host process may write. Impact is limited because the filename is derived from the input basename and always ends in .md; the agent typically already has filesystem write tools.
No shell/code injection or exfiltration found; residual risk is silent auto pip install (incl. --break-system-packages) and a caller-controlled output directory in the md_convert tool.
Rule: the dynamic audit can only raise risk, never lower it (raise-only, never an exemption); a wrong raise is withdrawn by human review and the tier falls back to the static verdict. Overall verdict = the stricter of the two layers.
Disclaimer: the static scan and the dynamic security audit are both automated analysis, not a security guarantee. Always review what you install.
Activity
Last commit 2026-10-04 · activity: Active
• Repo created: 2026-08-28
• Stars: ★ 2 · Forks: 0
• Health: Active — committed within last 30 days
Source
Author README
View on GitHub ↗The author's original README from the repository — fetched live, unedited (the English page loads the default README).