← ClaudeAtlas

aitk-scriptslisted

CLI entry point, bash scripts, sandbox scenarios, and lib functions. Use for `src/`, `manage-*.sh`, sandbox hooks, or shared `lib/` functions.
erclx/toolkit · ★ 1 · AI & Automation · score 74
Install: claude install-skill erclx/toolkit
# Scripts Read `docs/scripts.md` for structure, file inventory, and lib responsibilities before editing. ## Before editing - Read `.claude/GOV.md` before writing or editing code in `src/` or `scripts/`. ## Lib rules - Each lib file owns one concern. Read `docs/scripts.md` for responsibilities before adding or modifying. - Never duplicate logic that already exists in `lib/`. When adding a function, check if any existing script duplicates the logic and consolidate. ## Sandbox pattern - Each sandbox defines three hooks: `use_config` (flags before provisioning), `use_anchor` (remote repo as base), `stage_setup` (scenario state after provisioning). - Only `stage_setup` is required. End it with `log_step` describing what to run and what to expect. - Default behavior: no standards, no gov rules, no Gemini settings, auto-commit on. Declare only the flags you need in `use_config`. - For `claude/` scenarios, default to `SANDBOX_INJECT_SEEDS="true"`. See the rule and its two exceptions in `docs/sandbox.md`. - For multi-scenario scripts, call `select_or_route_scenario "Which scenario?" "a" "b"` instead of `select_option`. It reads `SANDBOX_SCENARIO` to skip the picker when set by `aitk sandbox <cat>:<cmd> <scenario>`. End every scenario `case` with a `*) log_error "Unknown scenario: $SELECTED_OPTION" ;;` arm. Use slug-style scenario names (no spaces) so agents can pass them without quoting. - One sandbox file per skill. Before adding a new scenario, check if `scripts/sandbox/<cat>