[blade-ai] --output silently ignores invalid formats / falls back to JSON

[dhruvraajeev]

What

Every blade-ai CLI command exposes --output/-o with help text json|yaml, but the value is an unvalidated str passed to format_output() in blade-ai/src/chaos_agent/cli/output.py:

if output_format == "yaml" and HAS_YAML:
    return yaml.dump(...)
else:
    return json.dumps(...)

So any unrecognized format silently returns JSON — --output xml, --output ymal (typo), etc. produce JSON with no error and no warning. The same silent fallback happens for --output yaml if PyYAML is somehow unavailable.

Affected commands (8)

config, capabilities_cmd, list_cmd, metric, confirm, inject, version, recover — each declares:

output: str = typer.Option("json", "--output", "-o", help="Output format: json|yaml")

Why it matters

A user scripting against the CLI who typos --output or requests an unsupported format gets JSON back silently, with no signal that their flag was ignored. For a tool whose output is consumed by automation, silently ignoring an invalid flag is a correctness problem.

Proposed fix

• Add an OutputFormat(str, Enum) in cli/output.py and type the --output option as that enum across all 8 commands, so Typer rejects invalid values at the CLI boundary and auto-generates accurate help.
• Make format_output() validate defensively: raise a clear ValueError on an unknown format, and a clear "install pyyaml" error when yaml is requested but PyYAML is unavailable (PyYAML is a declared dependency, so this only fires on a broken install).

Single, self-contained change (one helper + option typing + tests). Happy to submit a PR. (Observed on commit c571faa.)

Environment: blade-ai v0.3.0

[oss-sentinel-ai]

Issue Evaluation

Category: type/bug | Status: Under Review

Thank you for reporting this issue. It has been identified as a potential bug and is being reviewed.

Next Steps:

• The issue will be verified against the current codebase
• If confirmed, a fix proposal may be generated automatically
• You can reply /approve to proceed with PR generation once the spec is ready

---

Automated evaluation by github-manager-bot

[woyaonidish]

Thanks for the report — confirmed this is a real issue.

Verified in the current codebase:

1. format_output() in cli/output.py uses a plain if output_format == "yaml" and HAS_YAML / else branch — any unrecognized format (typos like ymal, unsupported formats like xml, or yaml when PyYAML is unavailable) silently falls back to JSON with no error or warning.

2. All 8 commands (config, capabilities, list, metric, confirm, inject, version, recover) declare --output as an unvalidated str via typer.Option, so Typer doesn't reject invalid values at the CLI boundary either.

For a CLI tool whose output is consumed by automation scripts, silently ignoring an invalid --output flag is a correctness problem — a user scripting against --output yaml would get JSON back and their YAML parser would fail with a confusing error instead of a clear "invalid format" or "PyYAML not installed" message.

We'll implement the proposed fix: add an OutputFormat(str, Enum) and type the --output option as that enum across all commands so Typer rejects invalid values upfront, plus make format_output() raise a clear error for unknown formats and missing PyYAML. Targeted for the next release.

Appreciate the detailed analysis — PR welcome if you'd like to get involved.