The Agent Skills Directory

[COMMAND_EXECUTION]: The skill defines several shell-based tasks for project analysis and maintenance.
Analysis Tools: The rules/project-analysis.md file specifies the use of cloc, pipreqs, npm, pydeps, and pyan for inspecting source code and analyzing dependencies.
Maintenance Commands: The rules/checkpoint.md file includes shell commands such as find .project-rebuilder/checkpoints -mtime +7 -delete and ls -t ... | xargs rm to manage state files.
[REMOTE_CODE_EXECUTION]: The RALPH (Recursive Autonomous Long-term Persistent Helper) mode described in rules/ralph-execution.md implements an autonomous task execution loop.
Execution Cycle: The agent autonomously acquires, executes, and verifies tasks, which inherently involves generating and running code to rebuild the project.
Autonomous Self-Healing: The skill includes logic to "try to fix" errors (尝试修复) up to two times, implying iterative code modification and execution without direct human oversight.
[PROMPT_INJECTION]: The skill has a significant attack surface for indirect prompt injection as it is designed to ingest and process external project data.
Ingestion Points: The skill retrieves and analyzes files from external project URLs and repositories (e.g., via the /rebuild analyze command).
Capability Inventory: The agent possesses file-writing capabilities and shell command execution access.
Boundary Markers: No explicit delimiters or security instructions are provided to the agent to distinguish between project data and its own system instructions, which could allow a reference project to override the agent's behavior.
Sanitization: There is no evidence of sanitization or validation of the external content before it is processed by the agent.

project-rebuilder