algal-research
Warn
Audited by Gen Agent Trust Hub on Sep 19, 2026
Risk Level: MEDIUMEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill uses
bunxto download and run thealgal-skillspackage from a public registry at runtime. - Evidence: The shell commands in
SKILL.mdinvokebunx algal-skillsto perform bundling and triage operations. - [REMOTE_CODE_EXECUTION]: The use of
bunxto execute an unversioned package (algal-skills) constitutes a remote code execution pattern, as it fetches and runs code from an external source without a fixed integrity check or version lock. - [INDIRECT_PROMPT_INJECTION]: The skill's primary function is to ingest data from up to twelve external URLs or source objects, which provides a surface for indirect prompt injection.
- Ingestion points: External URLs and source objects processed by the
research-bundlecommand inSKILL.md. - Boundary markers: The description mentions that sources are "clipped independently" and failures are isolated, but it does not specify the use of delimiters or instructions to prevent the agent from obeying instructions embedded in the clipped content.
- Capability inventory: The skill is explicitly allowed to use the
exectool and executes shell commands that interact with the local environment. - Sanitization: There is no evidence of sanitization or filtering applied to the remote content before it is bundled for the agent.
Audit Metadata