The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute local shell scripts (solo-research.sh, solo-dev.sh, solo-pipeline-status.sh) and system commands like tail and watch to manage and monitor pipelines.
Evidence: Steps 3b and Monitoring sections describe running these scripts with user-provided arguments.
[INDIRECT_PROMPT_INJECTION]: The skill exhibits an attack surface where untrusted user input is parsed and interpolated into subsequent skill invocations and shell commands.
Ingestion points: User-supplied $ARGUMENTS (e.g., project names, stack, features) are parsed in Step 1.
Boundary markers: The skill includes a mandatory 'Confirm with User' step (Step 2) that displays the intended actions and requires explicit approval before execution, which serves as a human-in-the-loop boundary.
Capability inventory: The skill uses Bash to run launcher scripts and Write to create project-local and global state files (e.g., ~/.solo/pipelines/).
Sanitization: There is no explicit evidence of programmatic sanitization or escaping of the arguments before they are passed to the Bash tool or interpolated into other skill calls.

solo-pipeline