create-ci-skill
Pass
Audited by Gen Agent Trust Hub on Nov 11, 2026
Risk Level: SAFEREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The skill downloads and installs executable code at runtime using the command
cargo install --git https://github.com/agoodway/goodflows.git --locked. This initiates a remote fetch and installation of a binary from a GitHub repository. - [COMMAND_EXECUTION]: The skill performs several shell-level operations, including querying git remote information (
git remote get-url origin), building code withcargo build, and executing third-party tool managers (npx skills add). - [EXTERNAL_DOWNLOADS]: The skill references and fetches external components from
https://github.com/agoodway/goodflows.gitand usesnpxto download and add additional skill modules (maintain-ci-skill) from theagoodway/goodflowsrepository. - [INDIRECT_PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by ingesting data from untrusted local configuration files (
goodflows.tomlor.goodflows.toml) and using that content to generate new instructions for an AI agent. - Ingestion points: The skill reads job names, commands, and workflow definitions from repository-local
goodflows.tomlfiles. - Boundary markers: There are no instructions to use delimiters or "ignore embedded instructions" warnings in the generated skill output to prevent the agent from obeying malicious instructions embedded in the config.
- Capability inventory: The generated skill is capable of executing shell commands through the
goodflowsCLI. - Sanitization: The skill lacks sanitization for the job commands and names extracted from the config file before they are written into the new
SKILL.mdfile.
Audit Metadata