The Agent Skills Directory

[COMMAND_EXECUTION]: Several scripts within the skill utilize subprocess.run to perform their tasks. governance_check.py orchestrates the audit by executing peer scripts like audit_project.py and audit_skills_repo.py. draft_issue.py invokes the standard GitHub CLI (gh) to create issues or questions based on audit findings. These executions are legitimate given the skill's purpose as a project auditor and workflow helper.
[INDIRECT_PROMPT_INJECTION]: The skill represents a surface for indirect prompt injection because it is designed to ingest and analyze untrusted external data (Kotlin source code, Gradle configurations, and Markdown documentation) from the projects it audits.
Ingestion points: Files are read by audit_skills_repo.py, classify_declarations.py, and audit_project.py (via governance_check.py).
Boundary markers: The scripts read file content directly without specific delimiters or warnings to the agent regarding potential embedded instructions.
Capability inventory: The skill has the ability to write to the file system (via draft_issue.py creating temporary files) and interact with external APIs (via the gh CLI for issue creation).
Sanitization: While draft_issue.py correctly uses shlex.quote to sanitize shell command arguments, the actual content of the generated issues is derived directly from the analyzed files without additional sanitization, which could lead to markdown-based injection in the final issue report.
[EXTERNAL_DOWNLOADS]: The documentation in references/governance-ci-enforcement.md provides instructions for integrating the audit into CI/CD pipelines by referencing a reusable GitHub workflow hosted in the author's own repository (ronjunevaldoz/kmp-agent-skills). This is a standard practice for managing shared CI workflows and originates from the recognized vendor/author.

kmp-audit