The Agent Skills Directory

Indirect Prompt Injection (LOW): The skill processes untrusted data from external sources (git logs and PR descriptions) which could contain malicious instructions designed to influence the agent's behavior during the 'learning' phase.
Ingestion points: The skill reads data from git log and gh pr list (GitHub CLI).
Boundary markers: Absent. There are no instructions or delimiters defined to separate untrusted data from the agent's core instructions.
Capability inventory: The skill utilizes Bash and Read tools, which allow for file system manipulation and command execution.
Sanitization: Absent. The skill performs frequency analysis and pattern normalization on raw text without escaping or validating the content.
Command Execution (SAFE): The use of bash for git, grep, sort, and uniq is appropriate for the skill's stated purpose and is restricted to the allowed-tools list.

compound-learnings