The Agent Skills Directory

COMMAND_EXECUTION (SAFE): The skill utilizes a bash orchestrator (migrate.sh) that executes a sequence of 10 migration scripts. These scripts perform necessary file system operations including directory creation (mkdir), file movement (mv), and content transformations using standard Unix utilities (grep, sed, awk, tail).
Evidence: The scripts are contained within the .claude/skills/migrate/scripts/ directory and are restricted to modifying project documentation and the .gitignore file.
PROMPT_INJECTION (SAFE): While the skill includes instructions that it is "mandatory" for the workflow, these are functional requirements for project consistency rather than attempts to bypass safety filters or extract system prompts.
DATA_EXFILTRATION (SAFE): Analysis of all 11 bash scripts confirmed the total absence of network-related commands such as curl, wget, or socket operations. The skill does not access sensitive user directories like ~/.ssh or ~/.aws.
DYNAMIC_EXECUTION (SAFE): The orchestrator script (migrate.sh) uses the source command to execute migration scripts in a loop based on the contents of the local migrations/ directory. This is a standard and safe implementation of a migration engine within a local package.
INDIRECT_PROMPT_INJECTION (SAFE): The scripts ingest and parse markdown files using regex and text processing tools. While these scripts do not sanitize every variable against YAML or shell injection, they use safe patterns (such as variable quoting and awk -v) that prevent the data within the documentation files from being executed as code.

migrate