add-stack-context

Pass

Audited by Gen Agent Trust Hub on Nov 10, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill is designed to discover and install external agent skills, MCP servers, and configuration files from vendor-controlled sources. It explicitly targets remote documents such as llms.txt to be used as live retrieval sources.
  • [COMMAND_EXECUTION]: The skill invokes external command-line tools including find-skills and the Skills CLI. It also applies vendor-provided installers and codemods to modify the project environment.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from remote vendor documentation and llms.txt files to generate instructions that are stored in persistent project files like AGENTS.md or CLAUDE.md, which will influence the behavior of future agent sessions.
  • Ingestion points: Remote llms.txt files, vendor documentation, and project manifests (package.json, lockfiles).
  • Boundary markers: The skill instructions recommend using "bounded managed instruction" blocks to separate vendor-provided content.
  • Capability inventory: File system write access (AGENTS.md, CLAUDE.md), shell command execution (Skills CLI, installers, codemods), and network access for content discovery.
  • Sanitization: The skill relies on verification of vendor ownership and official documentation rather than automated content sanitization or filtering of the retrieved instructions.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 10, 2026, 04:50 AM