The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands, specifically python3 plugins/siyu-execution/skills/siyu-pyq/scripts/pyq_lint.py, to validate the compliance of generated marketing copy. It also utilizes python3 tools/atoms_query.py to retrieve brand-specific data from an internal repository.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface where user-provided materials (such as brand details and product themes) are used to generate marketing copy that is then passed through a shell pipe. This creates a potential command injection vector if the generated output contains shell metacharacters that are interpreted by the host system.
Ingestion points: User inputs in SKILL.md (Product themes, targets) and assets/真实语料-待填.md (Real-world brand materials).
Boundary markers: The instructions lack specific delimiters or "ignore" directives to prevent the agent from obeying instructions embedded in the user-provided data during the copy generation or validation phase.
Capability inventory: Access to local file system scripts and the ability to execute subprocesses via the shell.
Sanitization: There is no explicit sanitization or escaping of the generated text performed before it is interpolated into the shell command sequence.
[COMMAND_EXECUTION]: The script scripts/pyq_lint.py performs dynamic path manipulation by modifying sys.path at runtime to locate and import internal project dependencies from the parent repository structure.

siyu-pyq