The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data that is subsequently processed by the autonomous Cline CLI agent, creating an injection surface.
Ingestion points: The prompt argument, --prompt-file, --session-file, and --output-schema inputs processed in scripts/run_cline.py.
Boundary markers: The build_prompt function in scripts/run_cline.py uses text-based headers (e.g., 'Role:', 'Execution constraint:', 'Prior conversation context:') to separate instructions from data.
Capability inventory: The skill invokes the cline CLI using subprocess.run. The cline agent possesses broad capabilities including file system access and shell command execution.
Sanitization: There is no evidence of sanitization or validation of the content within the ingested prompt or source files before they are passed to the CLI agent.
[COMMAND_EXECUTION]: The skill core functionality involves executing the local cline CLI through subprocess.run calls in scripts/run_cline.py. This is the intended behavior for this tool runner.
[PERSISTENCE]: As documented in references/runtime-reference.md and implemented in scripts/run_cline.py, the skill will mutate the user's global Cline configuration (~/.cline/data/settings/providers.json) when the --model flag is used without an isolating --data-dir. This persists changes to the provider's default model across future interactive sessions.

cline-runner