The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs agents to use uv run python -c to resolve project-specific paths by importing a local module dh_paths. This involves executing shell commands to determine environment context.\n- [DYNAMIC_EXECUTION]: Delegation prompts in references/agent-prompts.md utilize inline Python code to dynamically resolve the plan_dir variable at runtime.\n- [PRIVILEGE_ESCALATION]: The mock_binary fixture in SKILL.md programmatically sets executable permissions (0o755) on dynamically created files. This is a standard testing pattern but falls under privilege modification.\n- [INDIRECT_PROMPT_INJECTION]: The skill ingests data from external modernization plans to generate reports and tests, creating a surface for indirect instructions to influence agent behavior.\n
Ingestion points: The skill reads {plan_dir}/stinkysnake-plan.md in delegation prompts.\n
Boundary markers: No explicit delimiters are defined for the external content.\n
Capability inventory: Sub-agents are instructed to write files to the repository and execute shell commands for path resolution.\n
Sanitization: There is no evidence of sanitization or validation of the ingested plan content before it is used in sub-agent prompts.

python3-testing