Reversal — Agent Input Reliability Layer

PassAudited by ClawScan on May 11, 2026.

Overview

This appears to be a disclosed document and URL parsing integration, but it relies on external code/services and can process user-selected local files.

Install only if you trust the external Reversal repository and service. Prefer the local setup for sensitive documents, keep API keys protected, and ask the agent to parse only specific files or URLs you intend to share.

Findings (5)

Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.

What this means

Installing the MCP server could execute code from an external repository and package dependencies.

Why it was flagged

The skill tells users to install and run external code that is not included in the reviewed artifact set. This is disclosed and central to the skill, but users must trust that repository and its dependencies.

Skill content
git clone https://github.com/Etytabs/REVERSAL ... pip install -r requirements.txt ... pip install -e .
Recommendation

Review the repository and dependency files, use a virtual environment or container, and pin trusted versions before installing.

What this means

API keys could allow access to paid or private provider services if exposed or misused.

Why it was flagged

The skill discloses use of provider credentials for optional image parsing and remote Reversal API access. These credentials are expected for the stated integrations, but they are sensitive.

Skill content
Set `ANTHROPIC_API_KEY` only if you need image/dashboard screenshot parsing ... "Authorization": "Bearer sk-rev-YOUR_KEY"
Recommendation

Store keys in a secrets manager or environment-specific config, avoid committing them, and revoke or rotate them if exposed.

What this means

Content submitted for parsing may be processed by a third-party service rather than staying fully local.

Why it was flagged

The HTTP MCP configuration sends requests to a remote Reversal service. This is disclosed and purpose-aligned, but it creates an external data boundary for parsed URLs or documents.

Skill content
"transport": "http", "url": "https://api.reversal.dev/v1/mcp", "headers": { "Authorization": "Bearer sk-rev-YOUR_KEY" }
Recommendation

Use the local stdio option for sensitive files when possible, and review the provider’s privacy, retention, and access-control terms before sending confidential content.

What this means

Parsed content might be stored temporarily or persistently depending on the external implementation.

Why it was flagged

The security section mentions caching parsed JSON. This appears intended as safe implementation detail, but the artifact does not describe cache retention or isolation.

Skill content
JSON-only Redis cache — no pickle, no RCE surface
Recommendation

Check cache retention settings and avoid submitting highly sensitive documents unless storage behavior is understood.

What this means

If used carelessly, the agent could process a local file that contains private information.

Why it was flagged

The described tools can read or upload local files by path. That is aligned with document parsing, but it should remain limited to files the user intends to process.

Skill content
Input: source — URL (https://…) or file path (/path/to/file.pdf) ... upload_file(file_path)
Recommendation

Give the agent explicit file paths and avoid broad instructions such as parsing entire folders or unknown local files.