done-means-donelisted
Install: claude install-skill PyModel/done-means-done
# Done Means Done
Complete every authorized outcome. Fix and report every confirmed defect discovered in the authorized project. Verify the final integrated result. Preserve unfinished obligations across interruptions.
This is one skill backed by the included `dmd` CLI. `task.json` is the source of truth; reports and handoffs are generated views. Do not maintain a second independent checklist or write `COMPLETE` into state yourself.
## Runtime and activation
For a local Claude Code skill, the runtime is `python3 "${CLAUDE_SKILL_DIR}/bin/dmd"`; the host's skill substitution for this session is `${CLAUDE_SESSION_ID}`. These are skill-rendered substitutions, not a promise that shell environment variables exist. When another host leaves them literal, resolve the actual installed skill directory and host session identity. Never invent an ID or claim hook enforcement without a real binding.
Below, `dmd` means the included executable on `PATH`. Prefer the PATH shim: `python3 "${CLAUDE_SKILL_DIR}/hooks/install.py" --link-bin --no-hooks --apply` symlinks `bin/dmd` into `~/.local/bin` once, and every later shell and tool call finds it. When no shim is installed, the fallback is a shell function defined in the same command block that uses it: `dmd() { python3 "${CLAUDE_SKILL_DIR}/bin/dmd" "$@"; }`. Agent harnesses do not carry functions between tool calls, and a variable holding `python3 /path/dmd` is one word to zsh and fails silently when expanded. Read [commands](references/comm