The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides various shell script templates for automating developer workflows using the claude -p command. These are intended for user execution to chain agent tasks.
[EXTERNAL_DOWNLOADS]: Documents external tools such as continuous-claude. The skill includes explicit security warnings against piping remote scripts directly to the shell and advises users to review external code before installation.
[SAFE]: Identifies an indirect prompt injection surface inherent to autonomous loops. The architectures describe agents reading external specification files, RFCs, and CI logs which could contain untrusted data. Ingestion points include specs/, docs/, and gh run view output. The skill lacks explicit sanitization logic for these inputs, relying on the agent's core safety boundaries.

autonomous-loops