skills/oldwinter/skills/darwin-skill/Gen Agent Trust Hub

darwin-skill

Pass

Audited by Gen Agent Trust Hub on Oct 7, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill's primary purpose is to read, evaluate, and rewrite instructions from other skill files in the agent's environment. This ingestion of untrusted external instruction data into the agent's processing loop creates a surface for indirect prompt injection, where a targeted skill could influence the Darwin agent's behavior.
  • Ingestion points: Scans and reads all instruction files located in .claude/skills/*/SKILL.md.
  • Boundary markers: The skill does not use specific delimiters or instructions to isolate the content of target skills from its own optimization logic.
  • Capability inventory: The skill possesses capabilities to execute shell commands (git, npm, npx), write to the filesystem, and spawn independent sub-agents.
  • Sanitization: The skill performs no filtering or sanitization of the instruction text it ingests before using it in its optimization routines.
  • [EXTERNAL_DOWNLOADS]: The skill fetches necessary components and assets from well-known infrastructure providers.
  • Evidence: References a project archive for manual installation hosted on Cloudflare R2 (r2.dev) and downloads browser binaries for Playwright using standard NPM/npx commands.
  • [COMMAND_EXECUTION]: The skill performs shell-based operations essential for its version-controlled optimization workflow.
  • Evidence: Utilizes git to manage changes in instructions via commits and reverts, and executes npx playwright to generate high-definition visual reports from HTML templates.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 7, 2026, 01:53 PM