set-custom-rules

Warn

Audited by Gen Agent Trust Hub on May 14, 2026

Risk Level: MEDIUMEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill downloads the 'cc-safety-net' package from the public NPM registry at runtime via the 'npx' command. Evidence: 'npx -y cc-safety-net' in SKILL.md.
  • [REMOTE_CODE_EXECUTION]: The skill executes the downloaded 'cc-safety-net' package to retrieve schema documentation and verify user configurations. Evidence: Steps 1 and 8 in SKILL.md.
  • [COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands to interact with the system environment, including reading configuration files and running package commands. Evidence: Use of 'cat' and 'npx' commands in SKILL.md.
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it reads and processes data from external configuration files without validation. \n
  • Ingestion points: Reads '~/.cc-safety-net/config.json' and '.safety-net.json' in SKILL.md (Step 5). \n
  • Boundary markers: Absent; no delimiters or instructions are provided to the agent to treat the file content as untrusted. \n
  • Capability inventory: Shell execution via 'npx' (SKILL.md, Steps 1 and 8) and file writing (SKILL.md, Step 7). \n
  • Sanitization: Absent; the skill does not specify validation or schema enforcement of the JSON content before processing.
Audit Metadata
Risk Level
MEDIUM
Analyzed
May 14, 2026, 07:59 AM
Security Audit — agent-trust-hub — set-custom-rules