Borrowing it
Nothing to install: this file belongs to pvliesdonk/markdown-vault-mcp. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/pvliesdonk/markdown-vault-mcp/main/.agents/skills/writing-model-facing-text/SKILL.mdgit clone --depth 1 https://github.com/pvliesdonk/markdown-vault-mcpWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/pvliesdonk/markdown-vault-mcp/writing-model-facing-text)<a href="https://agentmods.dev/skills/pvliesdonk/markdown-vault-mcp/writing-model-facing-text"><img src="https://agentmods.dev/badge/skills/pvliesdonk/markdown-vault-mcp/writing-model-facing-text/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/pvliesdonk/markdown-vault-mcp/writing-model-facing-text"><img src="https://agentmods.dev/badge/skills/pvliesdonk/markdown-vault-mcp/writing-model-facing-text.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What 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 5.1 | $0.00056 | $0.03175 |
| Opus 5.5 | $0.00022 | $0.01270 |
| Sonnet 5.5 | $0.00011 | $0.00635 |
| Haiku 4.5 | $0.00006 | $0.00317 |
Grade A, and why
writing-model-facing-text 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 yesterday.
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.
Copies of this mod
2 near-identical copies found in the catalogue:
- writing-model-facing-text — 91% identical, 9 lines differ
- writing-model-facing-text — 91% identical, 9 lines differ
How it starts
The opening of the file, as written. The whole thing — 253 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Writing model-facing text
Every docstring and description= a FastMCP server registers is sent to
the client on the wire, and most of it reaches the model as a hint at one
moment: when it chooses a tool and fills in the arguments. Text the model
cannot act on at that moment is not harmless. It costs context on every
turn, Claude Code cuts each tool description and each server's
instructions at 2,048 characters and keeps the start, and a model told
about an operator command or a resource it cannot fetch acts on it anyway.
The evidence for every rule here is in
docs/design/reference/mcp-model-facing-text.md; read it before arguing
with a rule.
Who reads each surface
| Surface | Reader | Read when | What clients do with it |
|---|---|---|---|
| Tool description | the model | choosing a tool, filling arguments | put in the system prompt; Claude Code truncates at 2,048; tool search indexes it |
| Parameter description | the model | filling that one argument | property in the JSON schema; tool search indexes it |
Tool title, annotations |
the client | rendering a label, deciding whether to confirm | hints only; spec says clients must treat them as untrusted |
| Server instructions | the model | session start | system prompt; the one text Claude Code loads before any tool under tool search; truncated at 2,048; spec: should not duplicate tool descriptions |
| Resource description | a human in a picker, sometimes the model | browsing or attaching context | application-driven: the model may never see the list and often cannot fetch it |
| Prompt and prompt-argument descriptions | a human in a slash-command picker | picking a prompt | user-controlled |
| Tool result and error text | the model | after the call | the only surface that can say what to do next given the state |
| Skill served over MCP (SEP-2640) | the model, after the host lists its name and description and the model asks to load it | on demand, only on a host that implements the extension | skill:// resources; other clients see plain resources; hosts treat the body as untrusted data the user may refuse to load |
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.
- yesterday Changed f99480a46c1b
- 18d ago Changed · +1 lines a534e888b94f
- 26d ago Changed · +37 lines a93119c97e0e
- 27d ago Changed · +5 lines 4c2ca66ecbf6
- 28d ago First seen · 210 lines · 56 tokens per session scan A fef4bb1e0be3
writing-model-facing-text is a skill published in the GitHub repository pvliesdonk/markdown-vault-mcp (35 stars, last pushed today), licensed MIT. It adds 56 tokens to every session and 3,175 once invoked, about $0.0002 per session on Opus 5.5. 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 2026-09-25.
Other skills, from other repositories
docs-rag
Use when a project's markdown docs, specs, ADRs or decision logs need to be searchable by meaning from the agent, when a searchnotes tool (docs-rag-mcp) is connected, when its results carry labels like [semantic 0.71], [both], [exact match] or [superseded → ...], when it answers "No relevant results", or when a…
mnemonic
Search local indexed markdown knowledge bases. Use when the user asks to find notes, dig up a concept from personal docs, cross-reference ideas across wikis, or answer from indexed local files. Triggers on "look up in notes", "search my docs", "what did I write about", "find in my vault", "check my index", "retrieve…
llmwiki
Build and use a persistent, citation-traceable knowledge wiki from documents, research papers, notes, or project documentation. Use for llmwiki projects, retrieving reusable evidence for an agent task, or compiling a corpus for repeated questions. A one-off file read or live web search does not need a wiki.
siyuan-plugin-development
Official frontend plugin development workflow. After the user explicitly opts in, clarify missing requirements, confirm a short plan, implement in a managed copy, verify, and package. Installation and activation require separate authorization.
link-memory
Use after important user-approved decisions, when durable context should be proposed or reviewed, and for explicit Link memory lifecycle work: remember, recall, review, update, archive, restore, forget, or explain local memories through the CLI without requiring MCP.
link-ingest
Use when raw files are present, source pages look stale, or a user asks to ingest notes into Link; refresh source-backed wiki pages, propose memories, and validate updates through the CLI without MCP.