algal-digest

Pass

Audited by Gen Agent Trust Hub on Sep 19, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill uses the bunx command to download and execute the algal-skills package at runtime. This package serves as a vendor-provided tool to generate the digest.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted git metadata including commit messages and diffs which could potentially contain malicious instructions.
  • Ingestion points: Git repository status, history, and diff information are read by the tool in SKILL.md.
  • Boundary markers: The skill mentions that it produces a 'structured report' and a 'summary', but does not explicitly document the use of secure delimiters to separate untrusted git data from agent instructions.
  • Capability inventory: The skill is permitted to use the exec tool to run the digest generator.
  • Sanitization: No explicit sanitization or filtering of commit messages or other git metadata is described.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 19, 2026, 06:05 AM