The Agent Skills Directory

[CREDENTIALS_UNSAFE]: In Phase 6, the skill instructs the agent to create a Django superuser with a hardcoded password ('admin') through a shell pipe. This practice introduces high-risk default credentials that could be easily exploited if the command is executed on shared or production-accessible environments.
[COMMAND_EXECUTION]: The skill relies heavily on shell command execution to perform its primary functions, including environment verification, migration planning, and running various security and testing tools. This includes the use of Python one-liners (python -c) and multi-line Python scripts executed via heredocs (<< EOF) to access internal application settings and database metadata.
[PROMPT_INJECTION]: Phase 12 involves the agent reading and analyzing output from git diff, which represents an untrusted data source. A malicious actor could embed instructions within a code change that are then processed by the agent during the verification loop.
Ingestion points: The output of git diff is directly analyzed by the agent.
Boundary markers: No boundary markers or 'ignore' instructions are provided to separate the diff content from the agent's instructions.
Capability inventory: The skill environment has broad capabilities including shell execution, file system access, and network access via package managers.
Sanitization: No sanitization or validation of the diff output is performed; it is analyzed as raw text.

django-verification