The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill documents an installation procedure for a third-party tool using a high-risk command: curl -fsSL https://raw.githubusercontent.com/AnandChowdhary/continuous-claude/HEAD/install.sh | bash. This pattern executes arbitrary remote code from a non-trusted source without integrity verification or user review.
[COMMAND_EXECUTION]: The documentation provides multiple shell scripts (e.g., daily-dev.sh) that chain autonomous agent calls with filesystem modifications, Git operations, and build tool execution. These workflows provide a large attack surface for arbitrary code execution if the agent is manipulated.
[PROMPT_INJECTION]: The skill architectures for "Continuous Claude" and "RFC-Driven DAG" are vulnerable to indirect prompt injection.
Ingestion points: The agent reads from SHARED_TASK_NOTES.md, RFC documents, and PRD files in the local repository.
Boundary markers: No explicit boundary markers or safety delimiters are defined in the provided script examples or instructions to isolate untrusted file content.
Capability inventory: The loops possess capabilities for file writing, shell command execution via bash, and Git repository management through the gh CLI.
Sanitization: There is no mention of sanitizing external input or validating specification files before they are interpolated into the agent's prompt context.

autonomous-loops