The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides a framework for executing arbitrary shell commands specified in LOOP.md files.
Evidence: The run_shell function in scripts/loop_run.py executes commands using bash -c (on Windows) or /bin/sh -c (on POSIX).
Evidence: scripts/_common.py utilizes subprocess.run to execute commands and git operations required for the loop lifecycle.
[INDIRECT_PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection because it processes untrusted data from repository files and command outputs.
Ingestion points: The skill reads the verify and skip_when fields from LOOP.md specification files and captures the standard output and error of executed verification commands.
Boundary markers: The skill's instructions for dispatching fix workers do not define specific boundary markers or instructions to ignore potential commands embedded in the failing output.
Capability inventory: The skill has the ability to execute arbitrary shell commands and read/write files within the project directory.
Sanitization: There is no sanitization or validation of the shell commands extracted from LOOP.md before execution.
[DYNAMIC_EXECUTION]: The skill employs dynamic loading and runtime shell command generation.
Evidence: scripts/path_config.py uses importlib.import_module to load sibling helper modules dynamically.
Evidence: scripts/discussion_records.py dynamically modifies sys.path to resolve project-specific runtime scripts.

gsd-path-loop