The Agent Skills Directory

[COMMAND_EXECUTION]: The skill relies on the fluentci CLI to interact with the project environment, including running workflows (fluentci run), exporting configurations (fluentci project export), and inspecting modules. These operations naturally involve executing shell commands on the user's behalf.
[EXTERNAL_DOWNLOADS]: The documentation notes that certain commands, such as fluentci run setup-rust setup --help or fluentci run ci --help, may download modules, sources, or artifacts from the FluentCI registry or other repositories to provision the necessary build environments.
[REMOTE_CODE_EXECUTION]: By design, the tool executes workflow steps defined in TOML files, which can include shell command arrays or WASM modules. This functionality is intended for CI/CD automation and includes instructions on maintaining security, such as avoiding hardcoded secrets.
[INDIRECT_PROMPT_INJECTION]:
Ingestion points: The skill processes local project files including fluentci.toml, .fluentci/workflows/*.toml, and fluentci.lock to understand the workflow structure.
Boundary markers: No explicit markers are defined for the agent to separate file content from instructions, though the skill emphasizes inspecting generated YAML and notes manually.
Capability inventory: The tool can execute shell commands via the shell module, perform network requests for module downloads, and write exported CI manifests to the filesystem.
Sanitization: The instructions include security best practices, such as preventing unchecked JSON or log data from being written directly into sensitive environment variables like GITHUB_ENV or GITHUB_PATH.

fluentci-ci-workflows