vscode language-model-tool-descriptions.instructions.md

Guidance for writing descriptions of tools that an AI coding agent can choose and call.

In plain words
What is it for?
Use it when adding or reviewing model-facing tool descriptions, including their invocation rules, defaults, side effects, and retry guidance.
Why use it?
It helps descriptions state when a tool applies, what it changes, and what prerequisites or risks matter, so the agent is less likely to choose it incorrectly.

Instructions file for GitHub Copilot

Install

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.

agentmods
npx agentmods add instructions/microsoft/vscode/language-model-tool-descriptions
Clone the repo
git clone --depth 1 https://github.com/microsoft/vscode

Made for: GitHub Copilot.

Per session 283 This file is loaded in full into every session.
When invoked 283 The same file — it is already loaded in full.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

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.

ModelPer sessionOnce invoked
Fable 5 $0.00283 $0.00283
Opus 5 $0.00142 $0.00142
Sonnet 5 $0.00057 $0.00057
Haiku 4.5 $0.00028 $0.00028

Measured 2d ago against content hash b3168c9482c2, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

vscode language-model-tool-descriptions.instructions.md 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.

.github/instructions/language-model-tool-descriptions.instructions.md · 22 lines

What it actually says

Language model tool descriptions

Apply these rules only when adding or changing a model-facing tool description, such as modelDescription. They do not govern display names, user-facing messages, confirmation text, or routine schema property descriptions.

  • Begin with a direct imperative statement of the tool's capability.
  • When tool selection could be ambiguous, state the positive invocation criteria explicitly.
  • Place exclusions for likely near-neighbor requests immediately after the positive criteria.
  • Describe persistent, external, costly, or otherwise consequential side effects.
  • State meaningful defaults, cross-field requirements, and prerequisite tool calls that are not obvious from the input schema.
  • Put approval, cancellation, retry, repeat-call, and polling guidance last.
  • Do not repeat the complete input schema in prose.
  • Keep descriptions concise and scale their detail with the ambiguity and cost of incorrect selection.
  • Preserve terminology and formatting conventions used by related tools.
  • Reserve MUST, ONLY, CRITICAL, and direct "Do not" language for likely, consequential misuse.

When correctness depends on specific model-facing guidance, add a focused test for the required semantic clauses. Prefer checking a small set of meaningful phrases or conditions over snapshotting the entire description. Do not test routine descriptive prose.

Changes

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.

  1. 2d ago First seen · 22 lines · 283 tokens per session scan A b3168c9482c2

Subscribe to this mod's changes

vscode language-model-tool-descriptions.instructions.md is an instructions file published in the GitHub repository microsoft/vscode (190,384 stars, last pushed today), licensed MIT. It adds 283 tokens to every session, about $0.0014 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.