What causes "Skill content rejected by virus scan" in Significant-Gravitas/AutoGPT?

Remove flagged content (embedded binaries, EICAR strings, obfuscated payloads) from the skill body and retry.

Significant-Gravitas/AutoGPT · error · HTTPException

Skill content rejected by virus scan

Error message

Skill content rejected by virus scan

What it means

Skill-upload returns 400 when the ClamAV-backed virus scan of the skill content raises VirusDetectedError or VirusScanError during `store_user_skill`. Both are collapsed into the same generic client message and logged server-side (logger.warning '[skills] virus scan rejected uploaded skill') so scan details are not leaked. VirusScanError also fires when the scanner itself fails (unreachable ClamAV), not only on actual detections.

Source

Thrown at autogpt_platform/backend/backend/api/features/v1.py:2695

            detail=(
                "File is not a valid SKILL.md — expected YAML frontmatter with "
                "'name' and 'description' followed by a markdown body."
            ),
        )
    try:
        stored = await store_user_skill(
            user_id,
            name=parsed.name,
            description=parsed.description,
            body=parsed.body,
            triggers=list(parsed.triggers),
            version=parsed.version,
        )
    except SkillLimitError as exc:
        raise HTTPException(status_code=409, detail=str(exc))
    except (VirusDetectedError, VirusScanError) as exc:
        logger.warning("[skills] virus scan rejected uploaded skill: %s", exc)
        raise HTTPException(
            status_code=400, detail="Skill content rejected by virus scan"
        )
    except ValueError as exc:
        raise HTTPException(status_code=400, detail=str(exc))
    return CopilotSkillInfo(
        name=stored.name,
        description=stored.description,
        triggers=list(stored.triggers),
    )


@v1_router.get(
    path="/skills/{name}",
    summary="Read a single copilot skill with its full SKILL.md body",
    operation_id="readCopilotSkill",
    tags=["skills"],
    responses={404: {"description": "Skill not found"}},
    dependencies=[Security(requires_user)],

View on GitHub (pinned to 9c8bb5550f)

Solutions

Remove flagged content (embedded binaries, EICAR strings, obfuscated payloads) from the skill body and retry.
If ALL uploads fail, check ClamAV health: `docker compose ps clamav` / `echo PING | nc localhost 3310`, and inspect server logs for the '[skills] virus scan' warning with the underlying exception.
Operators: ensure the clamav container is up and reachable before diagnosing content issues.

Defensive patterns

Strategy: try-catch

Try / catch

try {
  await api.uploadSkill(content);
} catch (e) {
  if (e.status === 400 && /virus scan/i.test(e.detail)) {
    show('Content rejected by virus scan — remove embedded binaries/scripts.');
    return;
  }
  throw e;
}

Prevention

Keep skill bodies to plain markdown; avoid embedded binaries or base64 payloads.
Never use EICAR strings in real uploads — test with a dedicated sandbox flag instead.
If every upload is rejected, suspect ClamAV availability and check the service before blaming content.

When it happens

Trigger: POST upload of a SKILL.md whose body/frontmatter contains signatures ClamAV flags (EICAR test strings, embedded malicious payloads), or — for VirusScanError — uploads while the ClamAV service is down/misconfigured.

Common situations: Testing the upload pipeline with EICAR test files; embedding base64 blobs or scripts in skill bodies that trip heuristics; local dev stacks where the clamav docker service isn't started, making every upload fail with this 400.

Related errors

AI-assisted analysis of Significant-Gravitas/AutoGPT@9c8bb5550f (2026-08-14). Data as JSON: /api/errors/19b03bad679de4b8. Report an issue: GitHub.