agensi preflight

The problem

Submitting a skill to the Agensi marketplace often results in a 24 to 48 hour wait only to face rejection due to minor security flags or formatting errors. Catching leaked secrets, dangerous commands, or malformed metadata manually is error-prone and slows down your distribution cycle.

What it does

• Performs an 8-point security audit covering secrets, dangerous commands, and network access.
• Automatically strips junk files like .DS_Store, node_modules, and git history from the package.
• Validates SKILL.md frontmatter and normalizes folder structures for marketplace compatibility.
• Identifies potential prompt injection risks that could override host agent instructions.
• Generates a marketplace-ready .zip file with the correct internal hierarchy.

Frameworks & tools

Python 3, YAML frontmatter, shell scripts, and standard Agensi skill packaging formats.

Why this beats prompting it yourself

Standard LLM prompts often miss obscured security risks or subtle packaging requirements. This skill uses a dedicated scanner to find exact strings and patterns that trigger marketplace rejection, providing a definitive pre-submission verdict rather than a guess.

Use cases

• Auditing a newly developed skill for hardcoded API keys before public release.
• Cleaning up local development artifacts and junk files before zipping a package.
• Fixing "No SKILL.md at root" errors that cause automated ingestion failures.
• Optimizing marketplace listing descriptions and tags for better discoverability.

Known limitations

The scanner uses heuristics. Passing locally is a strong indicator of success but does not guarantee manual approval by Agensi reviewers.