The Agent Skills Directory

[PROMPT_INJECTION]: The skill features 'Code Review' and 'Debugging Detective' modes that require the agent to ingest and process code snippets or files provided by the user. This creates a surface for indirect prompt injection, where an attacker could embed malicious instructions within the code to manipulate the agent's behavior, such as exfiltrating data or modifying the learning log with false information.
Ingestion points: SKILL.md (Mode 2 and Mode 3) and user-supplied arguments for the scripts/run_tests.py utility.
Boundary markers: The instructions do not specify the use of strict delimiters or protective warnings to prevent the agent from obeying instructions embedded in the user-provided code.
Capability inventory: The agent is instructed to write to references/user-progress/learning_log.md and can execute the provided shell-based utility scripts.
Sanitization: No explicit sanitization or validation of the input code is described in the skill's instructions.
[COMMAND_EXECUTION]: The scripts/run_tests.py script executes shell commands to run testing frameworks like pytest, unittest, and jest. This involves using subprocess.run with parameters influenced by user input (the target file or directory).
Evidence: scripts/run_tests.py invokes shell commands to execute tests on user-specified targets, which is a necessary but high-risk capability in a code-execution environment.

code-mentor