Dead image checker
No broken images in your Markdown docs — checked automatically at session end
Full-repo scan of all .md and .mdx files for broken image references (). Covers relative paths and absolute paths resolved from public/ (Next.js convention). Ignores external URLs and data: URIs — no network requests. Skips node_modules, .git, .next and .claude directories. Non-blocking warning listing every broken reference so it can be fixed before pushing.
What does the Dead image checker hook do?
Dead image checker is a Claude Code Stop hook. It fires automatically at that lifecycle event — outside the model, so it can't be skipped or forgotten. No broken images in your Markdown docs — checked automatically at session end.
As a Stop hook it runs after the action, reacting to what just happened rather than blocking it. Because it is a deterministic Node.js script, it executes on every matching event without relying on the model to remember — the guarantee that makes agentic workflows safe to automate.
Use cases
- Catch broken image paths in README and docs after the agent moves or renames image files
- Validate Next.js public/ asset references in Markdown before pushing
- Prevent missing screenshots or diagrams from appearing as broken images in GitHub or documentation sites
Tags
settings.json fragment
{
"hooks": {
"Stop": [
{
"hooks": [
{
"command": "node $CLAUDE_PROJECT_DIR/.claude/hooks/stop-dead-image-checker.mjs",
"type": "command"
}
]
}
]
}
}Script · .claude/hooks/stop-dead-image-checker.mjs
#!/usr/bin/env node
// @hookstack stop-dead-image-checker
// Vérifie les images relatives cassées dans tous les fichiers Markdown du repo (Stop).
// Scan complet — couvre la dette existante, pas seulement les fichiers de la session.
// Gère les chemins relatifs ET les chemins absolus (résolus depuis public/).
// Purement Node.js (fs + path), sans réseau ni dépendance externe.
import { existsSync, readdirSync, readFileSync } from "node:fs";
import { dirname, join, resolve } from "node:path";
import { fileURLToPath } from "node:url";
const SKIP_DIRS = new Set([
"node_modules",
".git",
"dist",
"build",
".next",
"out",
".claude",
]);
// Capture  — uniquement les images (le ! est obligatoire)
const IMAGE_RE = /!\[([^\]]*)\]\(([^)]+)\)/g;
function stripCode(content) {
// Supprime les blocs de code clôturés (``` ou ~~~) — multiline
content = content.replace(/^```[\s\S]*?^```\s*$/gm, "");
content = content.replace(/^~~~[\s\S]*?^~~~\s*$/gm, "");
// Supprime les spans de code inline
content = content.replace(/`[^`\n]+`/g, "``");
return content;
}
function isExternal(src) {
return (
src.startsWith("http") || src.startsWith("data:") || src.startsWith("//")
);
}
function walkMd(dir, { readdir = readdirSync, exists = existsSync } = {}) {
if (!exists(dir)) return [];
const results = [];
for (const entry of readdir(dir, { withFileTypes: true })) {
if (SKIP_DIRS.has(entry.name)) continue;
const full = join(dir, entry.name);
if (entry.isDirectory()) results.push(...walkMd(full, { readdir, exists }));
else if (/\.mdx?$/.test(entry.name)) results.push(full);
}
return results;
}
export function run(
_input,
{
projectDir = process.env.CLAUDE_PROJECT_DIR ?? process.cwd(),
readFile = readFileSync,
exists = existsSync,
readdir = readdirSync,
} = {},
) {
const mdFiles = walkMd(projectDir, { readdir, exists });
if (!mdFiles.length) return null;
const broken = [];
for (const file of mdFiles) {
let content;
try {
content = readFile(file, "utf8");
} catch {
continue;
}
for (const [, , src] of stripCode(content).matchAll(IMAGE_RE)) {
if (isExternal(src)) continue;
let abs;
if (src.startsWith("/")) {
// Chemin absolu → résolu depuis public/ (convention Next.js et sites statiques)
abs = join(projectDir, "public", src);
} else {
abs = resolve(dirname(file), src);
}
if (!exists(abs)) {
broken.push(`${file.replace(`${projectDir}/`, "")} → ${src}`);
}
}
}
if (!broken.length) return null;
return {
message:
`[dead-image-checker] ${broken.length} broken image reference(s) across docs:\n` +
broken.map((b) => ` - ${b}`).join("\n") +
"\n",
};
}
/* v8 ignore next 5 */
if (process.argv[1] === fileURLToPath(import.meta.url)) {
const input = JSON.parse(readFileSync(0, "utf8"));
const result = run(input);
if (result) process.stderr.write(JSON.stringify(result));
}
Learn more
Related hooks
- Session summary generationAn automatic changelog of what the agent shipped
- Save compaction summary to logKeep a readable trail of every compaction
- Docs consistency reminderNo more READMEs telling two different stories
- Force implementation doc at stopNo code change ships without a doc/implementation/ trace