The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill dynamically generates shell scripts (scripts/setup_branch_db.sh and scripts/cleanup_branch_db.sh) and Elixir source code (modifications to config/runtime.exs and mix.exs) based on the project environment detection.
[PRIVILEGE_ESCALATION]: The skill uses chmod +x to grant execution permissions to the bash scripts it generates locally in the scripts/ directory.
[COMMAND_EXECUTION]: The skill executes multiple system commands to manage the development environment, including git, mix, psql, createdb, dropdb, and gh (GitHub CLI).
[INDIRECT_PROMPT_INJECTION]: The skill accepts a <branch-name> argument from the user which is interpolated into shell commands. Although the skill implements some sanitization within the generated scripts using shell parameter expansion, the initial git worktree add call uses the argument directly without validation.
Ingestion points: <branch-name> argument in SKILL.md.
Boundary markers: Absent in the initial command calls.
Capability inventory: File system writes, shell script execution, and interaction with database management tools.
Sanitization: Partial sanitization is applied during the manual fallback steps using shell string manipulation.
[CREDENTIALS_UNSAFE]: The generated database setup and cleanup scripts hardcode default PostgreSQL credentials (PGUSER="postgres", PGPASSWORD="postgres"). While common for local development, these are written to the user's filesystem as part of the automated workflow.

workie