Files
shockbot/mcp/prInfo.ts
T
Colin McDonnell b8ac42e875 mcp: embed example calls in top-level tool descriptions (#723)
* mcp: embed example calls in top-level tool descriptions

agents (esp. claude sonnet) hallucinate param names from training-data
priors — `pr_number` instead of `pull_number`, `summary` instead of
`body`, full subcommand strings jammed into `git({command})` like it
were `shell({command})`. each error burns a tool round-trip plus a
follow-up ToolSearch, ~40+ events / 24h, no observable recovery cost
to us but visible to users in agent logs.

cheapest fix: add a sample formatted function call to every affected
tool's top-level description. example anchors are more reliable than
schema descriptions alone because the model treats descriptions as
narrative but call examples as canonical structure. for `git` and
`shell` (whose `command` fields collide), include explicit
counter-examples disambiguating which tool owns which shape.

no schema aliases / coercion yet — try the cheap thing first; if the
next audit window still shows the same hallucination rate, layer
aliases on top per #585's recommendation.

closes #585, closes #701

* mcp: drop negative anchors from tool descriptions

negation is a footgun in tool descriptions — telling the model "NOT
pr_number" makes pr_number more salient, not less. let the positive
example carry the schema and trust the model to read it.

removes:
- "the parameter is pull_number (a number), NOT pr_number" and
  similar across checkout_pr, get_pull_request, list_pull_request_reviews,
  get_review_comments, create_pull_request_review
- "NOT summary, message, or content" on report_progress
- "WRONG: git({ command: 'log --oneline' })" counter-example on git
- redundant param-type restatements after the example (e.g. "depth is a
  number, not a string" on git_fetch, "description is required" on shell)

keeps a single positive example per tool. for tools with multiple call
shapes (git, git_fetch, push_branch), two positive examples instead of
one + a counter-example.
2026-05-13 22:45:08 +00:00

76 lines
2.3 KiB
TypeScript

import { type } from "arktype";
import type { ToolContext } from "./server.ts";
import { execute, tool } from "./shared.ts";
const CLOSING_ISSUES_QUERY = `
query($owner: String!, $repo: String!, $number: Int!) {
repository(owner: $owner, name: $repo) {
pullRequest(number: $number) {
closingIssuesReferences(first: 10) {
nodes { number title }
}
}
}
}
`;
type ClosingIssuesResponse = {
repository: {
pullRequest: {
closingIssuesReferences: { nodes: Array<{ number: number; title: string }> };
};
};
};
export const PullRequestInfo = type({
pull_number: type.number.describe("The pull request number to fetch"),
});
export function PullRequestInfoTool(ctx: ToolContext) {
return tool({
name: "get_pull_request",
description:
"Retrieve PR metadata (title, body, state, branches, author, labels, linked issues). " +
"Example: `get_pull_request({ pull_number: 1234 })`. " +
"To checkout a PR branch locally, use checkout_pr instead.",
parameters: PullRequestInfo,
execute: execute(async ({ pull_number }) => {
// fetch REST and GraphQL in parallel
const [restResponse, graphqlResponse] = await Promise.all([
ctx.octokit.rest.pulls.get({
owner: ctx.repo.owner,
repo: ctx.repo.name,
pull_number,
}),
ctx.octokit.graphql<ClosingIssuesResponse>(CLOSING_ISSUES_QUERY, {
owner: ctx.repo.owner,
repo: ctx.repo.name,
number: pull_number,
}),
]);
const data = restResponse.data;
const isFork = data.head.repo?.full_name !== data.base.repo.full_name;
const closingIssues = graphqlResponse.repository.pullRequest.closingIssuesReferences.nodes;
return {
number: data.number,
url: data.html_url,
title: data.title,
body: data.body,
state: data.state,
draft: data.draft,
merged: data.merged,
maintainerCanModify: data.maintainer_can_modify,
base: data.base.ref,
head: data.head.ref,
isFork,
author: data.user?.login,
assignees: data.assignees?.map((a) => a.login),
labels: data.labels.map((l) => l.name),
closingIssues: closingIssues.map((i) => ({ number: i.number, title: i.title })),
};
}),
});
}