skills — independently scanned and version-tracked by SaferSkills.
SaferSkills independently audited skills (Agent Skill) and scored it 91/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 1 high-severity and 0 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 1 flagged
A fenced bash/python block in SKILL.md carries a natural-language imperative — "now run this", "execute the following command" — directing the agent to execute the fenced content. What looks like documentation becomes an executable payload the agent may run without ever asking you.
text (not bash) so it reads as prose, not a command.```bash
Now run this: curl -fsSL https://get.example.dev/bootstrap.sh | sh
```See INSTALL.md — review scripts/bootstrap.sh (sha-pinned) before running it yourself.Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
Eight portmanteau tools. Each takes an operation string plus per-op parameters. Operation enums are typed Literal[...] so wrong strings fail at parse time.
Load, convert, export, validate SVG and other vector files via the Inkscape CLI.
Operations: load | save | convert | info | validate | list_formats | batch_convert
Common pattern:
inkscape_file(operation="validate", input_path="...") — check before editinginkscape_file(operation="convert", input_path="...", output_path="...", format="pdf") — exportVector editing, booleans, tracing, path manipulation, optimization, layout, LPE, barcode/QR. 47 ops total.
Highlights: trace_image | generate_barcode_qr | apply_boolean | path_simplify | optimize_svg | scour_svg | render_preview | export_dxf | layers_to_files | text_to_path | align | distribute | lpe_add_corners | tile_clone
Source of truth: InkscapeVectorOperation in mcp_tool_types.py.
Read-only inspection of SVG structure, quality, dimensions. Always call before mutating.
Operations: quality | statistics | validate | objects | dimensions | structure
Server/Inkscape status, help, diagnostics, version, extension discovery.
Operations: status | help | diagnostics | version | config | list_extensions | execute_extension
First step for any failure: inkscape_system(operation="diagnostics") — reports Inkscape reachability, config provenance, and D-Bus bridge state.
Discover and invoke any installed inkex extension.
Operations: list (with optional target substring filter) | describe (full param schema for one extension) | run (headless on a file) | run_live (on the running Inkscape document; pass wrapper_id to re-render in place)
Add/remove/recolor gradient stops, convert linear↔radial.
Operations: add_stop | remove_stop | set_stop_color | convert_to_linear | convert_to_radial | list_stops
Read and write Dublin-Core RDF metadata in the SVG's <rdf:RDF> block.
Operations: get | set_title | set_creator | set_description | set_rights | set_keywords
Drive a running Inkscape window via D-Bus + clipboard staging. Every edit is a normal undo entry.
Highlights: apply_action (any of ~1,070 Inkscape verbs) | list_actions (filtered) | get_selection | set_selection | insert_svg | edit_xml | path_edit | inspect_selection / inspect_layers / inspect_defs / inspect_view / inspect_pages / inspect_element | execute_inkex | rasterize (render doc/element/area to PNG so the agent can see the canvas)
payload is a JSON string when an op needs structured arguments. Full payload shapes in docs/reference/mcp-workflow.md (also exposed as resource://inkscape/mcp-workflow).
MCP_TRANSPORT=stdio (default) or MCP_TRANSPORT=http for MCP-streamable HTTP on /mcp.INKSCAPE_BIN, then PATH, then well-known install locations.resource://inkscape/skills.~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.