skills/cosmix/loom/loom-skills/Gen Agent Trust Hub

loom-skills

Pass

Audited by Gen Agent Trust Hub on Oct 18, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and execute instructions from external files, creating a potential attack surface.
  • Ingestion points: The agent is instructed to use the Read tool to fetch content from ~/.claude/loom-skill-catalog/<argument>/SKILL.md and ~/.claude/skills/<name>/SKILL.md based on user-provided arguments.
  • Boundary markers: There are no instructions provided to treat the ingested content as untrusted data or to ignore embedded instructions within the fetched files.
  • Capability inventory: The skill uses the Read tool to fetch content and subsequently directs the agent to "follow it as if it had been loaded directly," effectively executing the instructions contained within the external files.
  • Sanitization: The skill does not implement any validation or sanitization of the content fetched from the local filesystem before following its instructions.
  • [DYNAMIC_EXECUTION]: The skill implements a dynamic instruction loading mechanism.
  • The instruction to "follow it as if it had been loaded directly" allows the agent to adopt new behavioral guidelines at runtime based on the contents of files retrieved via the Read tool.
  • [DATA_EXPOSURE]: The skill directs the agent to access the ~/.claude/ directory.
  • This directory is part of the agent's sensitive configuration environment, containing other skills and potentially session data or configuration files.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 18, 2026, 02:02 PM