The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to use standard development tools such as git (for branch management and commits), gh (for GitHub issue and PR management), and project-specific build/test commands. These operations are within the expected scope for a development-focused agent skill.
[REMOTE_CODE_EXECUTION]: The skill mentions external drivers and fresh-context workers in references/BATCH_AND_PORTABILITY.md, but these are documented as architectural recommendations for the host environment and do not involve the skill downloading or executing untrusted code directly.
[PRIVILEGE_ESCALATION]: No privilege escalation patterns like sudo or modification of sensitive system files were found. All file operations are targeted at the local repository and specific documentation paths.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data (PR descriptions, issue bodies, and repository state). It implements boundary markers such as fixed schemas for progress.md and mandatory review gates (/review-change) to mitigate the risks associated with processing untrusted data.
[DYNAMIC_EXECUTION]: The skill uses node scripts/pre-execution-snapshot.mjs to verify plan snapshots. This is a local script used for integrity verification and does not involve dynamic execution of untrusted input.
[CREDENTIALS_SAFE]: No hardcoded credentials or access to sensitive credential stores (e.g., ~/.aws/credentials, .env) were detected. The use of the gh CLI assumes the agent is already authenticated in the environment.

execute-phase