Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add skills/stencila/stencila/conversionnpx skills add stencila/stencila --skill conversiongit clone --depth 1 https://github.com/stencila/stencilaWhat it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00081 | $0.00899 |
| Opus 5 | $0.00041 | $0.00449 |
| Sonnet 5 | $0.00016 | $0.00180 |
| Haiku 4.5 | $0.00008 | $0.00090 |
Grade A, and why
conversion scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 82 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Converting Stencila documents
NO_COLOR=1 stencila convert input.docx output.smd --yes
Formats are inferred from file extensions; --from/--to override. The
input may be a path, a URL, or a bare identifier (see Ingestion below).
stencila formats list prints the live capability table — which formats
decode (From), encode (To), and are lossless. Consult it rather than
guessing.
Lossless vs lossy
Only the data formats round-trip losslessly: CBOR, CBOR+Zstd, JSON, JSON+Zip, JSON5, JSON-LD, YAML. Everything else — DOCX, PDF, LaTeX, JATS, HTML, Markdown flavours, IPYNB — is lossy to some degree.
When converting to a lossy target, quantify what is dropped:
--input-losses— action on losses decoding the input--output-losses— action on losses encoding the output
Both accept ignore, trace, debug (default), info, warn, error,
abort, or a .json/.yaml filename to write the losses to.
After a lossy conversion, tell the user what was lost rather than silently
producing a smaller document.
Decoding options that matter
--fine/--coarse— decoding granularity. Fine (the default for most formats) decodes to the finest structure; coarse preserves un-modelled markup in larger blocks and is the default for LaTeX. Use--coarsewhen fine decoding of a partially supported format mangles structure.--pages N-M/--exclude-pages— page selection for multi-page inputs (e.g. PDFs):--pages 1,3,5-7,--pages 2-, keywordsodd/even.--include-structuring/--exclude-structuring— structuring operations that infer document structure (sections, abstracts, references) from loosely structured input. Example:--include-structuring sections-to-abstract.--ignore-artifacts— re-download / re-process instead of using cached intermediate artifacts in.stencila/artifacts/.
Ingesting published papers
convert accepts identifiers directly as the input — a DOI, arXiv id, or
PubMed Central id — and fetches the document:
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 2d ago First seen · 82 lines · 81 tokens per session scan A 275545aa84f7
conversion is a skill published in the GitHub repository stencila/stencila (900 stars, last pushed 3d ago), licensed Apache-2.0. It adds 81 tokens to every session and 899 once invoked, about $0.0004 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
pdf-processing
Extract text from PDFs, fill forms, and merge documents.
wegent-knowledge
Knowledge base management and search tools for Wegent. Provides capabilities to list, create, update, and search knowledge bases and documents using RAG retrieval. Use this skill when the user wants to manage knowledge bases, documents, or search for information programmatically.
document-skills/pdf
Comprehensive PDF manipulation toolkit for extracting text and tables, creating new PDFs, merging/splitting documents, and handling forms. When Claude needs to fill in a PDF form or programmatically process, generate, or analyze PDF documents at scale.
engine-concurrency
Use when adding, reviewing, or debugging parallelism in ArcadeDB engine or server code - covers the dedicated thread pool inventory and sizing, saturation policy, lock-free read patterns, locking rules, Micrometer pool metrics, and the checklist a new pool must satisfy. Also use when deciding where to submit forked…
dingtalk-docs
Access DingTalk documents and knowledge spaces through the DingTalk document MCP. Use this skill whenever the user selects a DingTalk knowledge space, folder, or document.
code-janitor
Reviews, cleans up, and verifies code as production-ready in a folder/package or recent git changes. Groups files by intent, then runs a 5-chain pipeline per group (Simplification → Abstraction → Readability → Documentation → Production-Readiness Sweep) with one supervisor granting an industry-standard…