Codifica

• Analyzed: May 12, 2026, 4:28 AM
• Findings: 1
• Version: 1.0.1
• Hash: bcc8b9d325d556541cdeb7e89f25c99a61ec82f631dade4e4990f098aefd3083
• Source: palm
• External report: View on VirusTotal

Findings (1)

The skill defines a protocol that heavily relies on the AI agent reading and interpreting various user-controlled files within the repository (e.g., `codifica.json`, `codifica-spec.md`, `state` files, `context.files`, `artifacts` as detailed in `SKILL.md`). While the skill's stated purpose is benign (protocol for agent coordination), this extensive instruction to read arbitrary text from user-controlled files creates a significant prompt injection vulnerability. An attacker who can modify these files could potentially inject new instructions to the agent, leading to unintended actions. This is a vulnerability that allows attacks, rather than being an attack itself, hence 'suspicious' instead of 'malicious'.