The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data in the form of C++ source files (*.cpp, *.h, *.hpp) and git diff output to perform its review. This content is provided directly to AI subagents, creating a surface for potential injection attacks where malicious code comments could attempt to influence the agent's behavior.
Ingestion points: Source files identified through globbing and output from git diff commands as specified in SKILL.md.
Boundary markers: The instructions lack specific delimiters or "ignore embedded instructions" warnings when providing the source code context to the six analysis subagents.
Capability inventory: The subagents are granted read access to the codebase and symbol search capabilities. The main agent can execute a local Python script and generate formatted reports.
Sanitization: No sanitization, escaping, or filtering of the source code content is performed before it is interpolated into the agent prompts.
[COMMAND_EXECUTION]: The skill executes shell commands including git diff to retrieve code changes and python3 references/lint-scripts/qt_review_lint.py to run a static analysis script. These commands are localized to the project environment and are used for their intended purpose of code auditing.

qt-cpp-review