The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the Bash subagent and various file-system tools to perform structural changes and validate the refactor. It specifically calls git mv, mkdir, and executes local test suites like pytest or npm test. While these operations are consistent with the skill's primary purpose, they involve executing and manipulating code found within the target repository.
[INDIRECT_PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface (Category 8) because it ingests data from the local repository environment to drive automated planning and execution phases.
Ingestion points: The Explore subagent scans repository directories and reads configuration files such as pyproject.toml and package.json in SKILL.md (Phase 1).
Boundary markers: External context from previous steps is interpolated into prompts for the Plan and general-purpose subagents (e.g., {analysis_results}, {migration_plan}) without the use of delimiters or instructions to ignore embedded commands.
Capability inventory: The skill possesses the ability to modify the file system via git, update import statements in source files, and execute arbitrary shell commands via the Bash subagent during the validation phase.
Sanitization: No sanitization or validation of the ingested repository metadata (file names, module names, or config values) is evident before the data is used to generate the migration tasks.

discipline-refactor