The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and act upon complex data provided in YAML formats from external roles such as orchestrators or specialists.
Ingestion points: Data is received via the task, scoped_task (aliased as junior_task), math_handoff, and stats_handoff blocks as defined in SKILL.md.
Boundary markers: The skill utilizes structured YAML blocks for input, but it lacks specific instructions or delimiters to ensure the agent ignores potentially malicious natural language instructions embedded within these data fields.
Capability inventory: The skill has access to the Bash tool for executing commands (ruff, mypy, pytest, coverage, git) and the Write tool for creating or modifying files in the codebase and the /tmp directory.
Sanitization: There are no instructions for sanitizing, validating, or escaping the external content before it is processed or used to generate code.
[COMMAND_EXECUTION]: The skill utilizes shell commands to perform standard development tasks, which could be exploited if the agent is misled by malicious input.
Evidence: The skill explicitly instructs the use of Bash to run ruff check, mypy --strict, pytest, coverage, and git commands.

python-developer