The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to Indirect Prompt Injection (Category 8) because it ingests data from untrusted external files to generate or migrate QA checklists.
Ingestion points: SKILL.md (Mode: Migrate, Mode: Audit) instructs the agent to read QA.md, CLAUDE.md, README.md, ROADMAP.md, and QA_BACKLOG.md from the project environment.
Boundary markers: The skill does not define specific boundary markers or use 'ignore embedded instructions' warnings when parsing content from these files.
Capability inventory: The skill has the authority to write files (QA.md) and interact with the user. It also identifies and stores shell-based 'Setup' commands.
Sanitization: No sanitization or escaping of extracted content is specified before the data is processed or written back to the checklist.
[COMMAND_EXECUTION]: The skill manages a 'Setup' command (e.g., npm run dev) that is stored in the QA.md file and used to prepare the environment for testing. While the instructions suggest user interaction, the requirement to 'Verify the Setup command runs' before writing the file could lead an agent to execute arbitrary shell commands if it encounters a malicious project configuration.

track-qa