Generated commit message failed validation: ${generated.vali

Error message

Generated commit message failed validation: ${generated.validationError}

What it means

runLegacyCommitCommand validates the model-generated commit message before committing. If the generated message fails validation (generated.validationError is set) in the non-interactive path, it throws instead of writing an invalid commit. The interactive path only warns; this throw happens when there is no user to correct the message.

Source

Thrown at packages/coding-agent/src/commit/pipeline.ts:55

				return;
			}
			process.stderr.write("No changes to commit.\n");
			return;
		}
		throw error;
	}

	const commitMessage = formatConventionalCommit(generated.commit);
	if (args.dryRun) {
		process.stdout.write("\nGenerated commit message:\n");
		process.stdout.write(`${commitMessage}\n`);
		if (generated.validationError) {
			process.stderr.write(`Warning: generated message requires manual correction: ${generated.validationError}\n`);
		}
		return;
	}
	if (generated.validationError) {
		throw new Error(`Generated commit message failed validation: ${generated.validationError}`);
	}

	if (!args.noChangelog) await updateChangelog(cwd, args);
	try {
		await vcs.requireGit(cwd).commitCreate(commitMessage, {});
	} catch (error) {
		if (vcs.isVcsError(error)) abortOnGitFailure("Commit failed", error);
		throw error;
	}
	process.stdout.write("Commit created.\n");
	if (args.push) await pushOrAbort(cwd);
}

async function updateChangelog(cwd: string, args: CommitCommandArgs): Promise<void> {
	const settings = await Settings.init({ cwd });
	const authStorage = await discoverAuthStorage();
	const registry = new ModelRegistry(authStorage);
	await registry.refresh();

View on GitHub (pinned to 9690622007)

Solutions

Re-run commit generation with a stronger model or regenerate until validation passes.
Write the commit message manually and commit with git directly.
Inspect validationError text and relax/adjust project commit-message conventions if overly strict.
Adjust prompting/temperature settings used by the commit pipeline if consistently failing.

Example fix

// before
omp commit   // throws: Generated commit message failed validation: subject exceeds 72 chars
// after
git commit -m "fix: trim subject to fit conventional limits"

Defensive patterns

Strategy: try-catch

Validate before calling

const SUBJECT_MAX = 72;
function isValidCommitMessage(msg: string): boolean {
  const subject = msg.split("\n")[0];
  return /^(feat|fix|docs|style|refactor|perf|test|build|ci|chore|revert)(\(.+\))?!?: /.test(subject) && subject.length <= SUBJECT_MAX;
}

Type guard

function isGeneratedOk(g: { message: string; validationError?: string | null }): g is { message: string; validationError: undefined } {
  return !g.validationError;
}

Try / catch

try {
  await runLegacyCommitCommand(args);
} catch (err) {
  if (err instanceof Error && err.message.startsWith("Generated commit message failed validation")) {
    process.stderr.write(`${err.message}\nFalling back to manual commit.\n`);
  } else throw err;
}

Prevention

Use a model known to follow format instructions for commit generation.
Keep project commit conventions (types, subject length) documented where the pipeline prompt reflects them.
Regenerate or manually edit when the warning variant appears interactively instead of letting it reach non-interactive runs.
Pre-commit commitlint rules should match what the generator validates against to avoid repeated failures.

When it happens

Trigger: Running the commit command non-interactively when the LLM output violates message rules — e.g. subject too long, wrong conventional-commit type, multi-line subject, forbidden characters — so the validator sets validationError.

Common situations: Small/weak models producing malformed conventional-commit subjects; hook scripts or bots invoking the commit command in CI where nobody can edit the message; strict project commit lint rules the model doesn't satisfy.

Related errors

AI-assisted analysis of can1357/oh-my-pi@9690622007 (2026-08-31). Data as JSON: /api/errors/28da404ab389b75b. Report an issue: GitHub.