The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes various build and test commands discovered within the repository to verify project readiness. This includes running tools like pytest, jest, go test, and pnpm build in a dry-run or help mode to ensure they are properly configured.
[EXTERNAL_DOWNLOADS]: The remediation templates suggest adding pre-commit hooks that reference public repositories from well-known sources like the official pre-commit hooks and the Astral-sh (Ruff) repository.
[DATA_EXPOSURE]: The assessment logic scans for the presence of environment variable templates (.env.example) and verifies that sensitive files like .env are correctly ignored by version control. This is a standard security best practice and does not involve exfiltration.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from the repository (README files, source code, and configuration files) to generate its reports and remediation suggestions.
Ingestion points: Project files including README.md, source code (for environment variable detection), and build configuration files in workflow.md and remediation.md.
Boundary markers: No explicit boundary markers or instructions to ignore embedded prompts are defined for the sub-tasks.
Capability inventory: The skill can execute shell commands for verification purposes and write configuration/documentation files to the local filesystem.
Sanitization: No explicit sanitization of ingested file content is documented before it is processed by the scouts.

flux-opencode-prime