The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface where untrusted data from the codebase could influence agent behavior. If a source file contains adversarial instructions disguised as comments or data, the agent might inadvertently execute them while performing analysis or fixes.
Ingestion points: Reads local files and directories provided as arguments (e.g., src/, src/utils.py) or the entire project.
Boundary markers: Absent; the instructions do not provide delimiters or warnings to ignore embedded commands within the analyzed code.
Capability inventory: Includes file modification (Fix Issues) and shell command execution (ruff, pytest).
Sanitization: Absent; the skill does not explicitly define sanitization or validation steps for content ingested from the codebase.
[COMMAND_EXECUTION]: The skill workflow involves the execution of external tools via the shell.
Evidence: The 'Verify' step instructs the agent to run ruff check . and pytest to ensure code quality and correctness.

techdebt