lint-rule-authoring
Pass
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data from GitHub pull requests and repositories to evaluate lint rules, creating an attack surface for indirect prompt injection.
- Ingestion points: The skill reads pull request metadata via
gh pr viewand clones external repositories to perform parity runs, as documented inBASE-HEAD-PARITY.mdandLINT-RULES.md. - Capability inventory: The agent has access to shell commands (
gh,git,node,jq) and the Vercel Sandbox API for executing evaluators and managing cloud resources. - Sanitization: The skill implements security checks for
rootDirinputs, requiring them to be relative, normalized, and free of shell metacharacters like backticks, semicolons, and pipes. - Boundary markers: No specific delimiters or warnings for separating ingested content from instructions are defined in the skill's operational rules.
- [COMMAND_EXECUTION]: The skill instructs the agent to use various CLI tools including
gh,git,jq, andnodeto manage the lint rule evaluation lifecycle, including repository manipulation and script execution for differential runs.
Audit Metadata