The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to analyze project code and web resources to generate test files, which creates a vulnerability to instructions embedded in that data.
Ingestion points: The skill uses Read, Grep, and Glob to process local source code (SKILL.md) and WebSearch to retrieve information from the internet, both of which are attacker-controllable ingestion points.
Boundary markers: The instructions lack explicit delimiters or safety prompts to ensure the agent ignores natural language instructions found within the files it reads.
Capability inventory: The skill is permitted to use Bash for command execution and Write/Edit for modifying the project structure, which could be abused if the agent follows instructions injected into the source code.
Sanitization: There is no evidence of sanitization or filtering of external data before it is processed by the agent to create new test scripts or run shell commands.

pytest-coder