← ClaudeAtlas

lgtm-cli-controllisted

lgtm local CLI/TUI control skill. Use only when a selected PLAN.md phase changes CLI/TUI behavior, terminal output, prompts, interrupts, hangs, resize behavior, or terminal demos and needs repeatable local evidence.
yarlson/lgtm · ★ 4 · AI & Automation · score 70
Install: claude install-skill yarlson/lgtm
# lgtm CLI Control Use only when selected phase need user-visible CLI/TUI verify. Goal: repeatable local harness, not manual poke. ## Workflow 1. ID command, workspace, behavior under test. 2. Prefer repo-native harness: - integration tests - e2e tests - demo scripts - PTY helpers - expect scripts 3. No harness? Use temp local harness under `/tmp`. 4. Drive one action at a time. Wait for concrete output before next. 5. Capture smallest transcript that prove/disprove behavior. 6. Clean up temp sessions, processes, artifacts unless user say keep. 7. Turn findings into selected-phase fix or explicit blocker. ## Harness Options Prefer repo-native tools. If need: - `tmux` for managed terminal sessions - short PTY script for deterministic waits - existing runtime profilers for startup, hangs, memory No add testing dependency for one-off probe unless selected phase require. ## Guardrails No send credentials or destructive commands into harness. No hardcode paths from another repo. No keep harness code in repo unless selected phase need reusable test. No treat screenshots/transcripts as enough when stable automated test practical. ## Completion Criteria CLI control done when CLI/TUI behavior verified with local evidence, fixed within selected-phase scope, or blocked with clear reason.