The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The script scripts/initiative_registry.py delegates functionality to external Python scripts (such as registry_admin.py, git_workflow.py, compile_registry.py, and validate_registry.py) located within a user-provided repository checkout. This execution occurs without verifying the integrity or content of the scripts.
[DYNAMIC_EXECUTION]: The delegate function in scripts/initiative_registry.py uses subprocess.run to dynamically execute Python scripts. The specific script to be run is determined by the operation argument, and the execution path is derived from the --registry-path argument supplied at runtime.
[COMMAND_EXECUTION]: The skill uses the subprocess module to execute system commands (git) and spawn new Python processes. It passes arbitrary arguments collected from the user via argparse.REMAINDER to these processes.
[INDIRECT_PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection as it is designed to ingest and act upon data from external sources.
Ingestion points: Reads AGENTS.md, protocol specifications, and registry manifests from a remote GitHub repository (referenced in SKILL.md).
Boundary markers: The skill does not implement delimiters or explicit 'ignore embedded instructions' warnings for the data it processes.
Capability inventory: The skill has the capability to execute shell commands and Python scripts via the subprocess module in scripts/initiative_registry.py.
Sanitization: There is no evidence of validation, filtering, or escaping of the content retrieved from the external registry before it is used or processed.

zj-initiative-registry