fable-advisor

Pass

Audited by Gen Agent Trust Hub on Jul 30, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The script scripts/ask_fable.mjs uses spawnSync to execute the claude CLI tool. It passes user-provided data from standard input directly to this process using stdio: "inherit".
  • [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface.
  • Ingestion points: The skill ingests untrusted data from standard input, as demonstrated in the execution instructions in SKILL.md (node scripts/ask_fable.mjs <<'REVIEW' ...).
  • Boundary markers: No delimiters or "ignore embedded instructions" warnings are used when passing the input to the sub-agent.
  • Capability inventory: The claude sub-agent is explicitly initialized with file-system access tools (Read, Grep, Glob), which could be abused if the input contains malicious instructions.
  • Sanitization: There is no evidence of input validation, escaping, or filtering before the data is processed by the AI model.
  • [COMMAND_EXECUTION]: The script attempts to read a configuration file from a path outside of the skill's own directory structure (../../../claude-agents/fable-advisor.md), which creates a dependency on a specific local environment layout.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 30, 2026, 10:41 AM
Security Audit — agent-trust-hub — fable-advisor