The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill analyzes local project manifest files (e.g., package.json, pyproject.toml, go.mod) and module structures, creating an attack surface for indirect prompt injection via untrusted project metadata. While no explicit boundary markers or sanitization are specified for this data ingestion, the risk is mitigated by the mandatory approval-gated workflow (Analyze, Design, Plan, Execute) which requires human review at every stage. The skill's capabilities include git operations, file system writes, and execution of test scripts.
[COMMAND_EXECUTION]: Standard git operations such as fetch, pull, and checkout are utilized to maintain project synchronization and manage feature branches safely. these commands are restricted to the local environment and are consistent with professional development workflows.
[EXTERNAL_DOWNLOADS]: The skill references and recommends industry-standard CLI libraries (e.g., click, commander, cobra, clap) and testing tools from established organizations and public registries. These external resources are well-known, reputable components of modern software development ecosystems and are documented neutrally.

cli-builder