follow-graph-spec

Pass

Audited by Gen Agent Trust Hub on Aug 15, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands using npx graph-spec-cli, which is the primary tool for analyzing graphspec bundles. These commands are used to validate bundles, check coverage, and determine build order.\n- [EXTERNAL_DOWNLOADS]: The skill relies on npx graph-spec-cli, which downloads the tool from the public NPM registry at runtime. As the tool is essential for the skill's purpose and provided by the vendor, this is considered legitimate behavior.\n- [INDIRECT_PROMPT_INJECTION]: The skill creates an attack surface by reading and processing external graphspec bundles provided by the user.\n
  • Ingestion points: The agent is instructed to read markdown files and their frontmatter from a user-defined directory (Step 4, 5).\n
  • Boundary markers: While the skill includes a validate --strict step to verify bundle integrity, it does not suggest explicit delimiters or warnings to ignore malicious instructions within the natural language content of the specs.\n
  • Capability inventory: The agent is capable of command execution and code generation based on the spec content.\n
  • Sanitization: There is no evidence of content sanitization or filtering for the natural language text within the spec files.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 15, 2026, 08:39 PM
Security Audit — agent-trust-hub — follow-graph-spec