// Explain what livespec does and route the user to the right sub-command (seed, propose-change, critique, revise, doctor, prune-history). Invoked by /livespec:help, "what can livespec do", or when the user asks for an overview of livespec capabilities.
[HINT] Download the complete skill directory including SKILL.md and all related files
| name | help |
| description | Explain what livespec does and route the user to the right sub-command (seed, propose-change, critique, revise, doctor, prune-history). Invoked by /livespec:help, "what can livespec do", or when the user asks for an overview of livespec capabilities. |
| allowed-tools | Read |
Explain what livespec does at a high level and route the user to
the right sub-command for their intent. There is no Python
wrapper for help — the response is composed by the LLM in
narration, no bin/*.py invocation, no JSON.
/livespec:help, says "what can livespec do",
"how do I use livespec", or asks for an overview of livespec
capabilities.Brief overview. Explain in 1-2 sentences: "livespec is an LLM-mediated specification toolchain. You author natural-language specifications, and livespec's sub-commands govern the workflow: seeding the initial spec, filing proposed changes, critiquing, revising into version snapshots, and running structural-invariant checks."
Route to the right sub-command. Map the user's goal to one of:
/livespec:seed) — start a brand-new spec in
an empty project. The skill walks a 3-question pre-seed
dialogue (template, sub-specs?, intent), then generates
and writes the initial SPECIFICATION/ tree plus
auto-captured seed proposed-change./livespec:propose-change) — file a
proposed change against an existing spec. Lands at
<spec-target>/proposed_changes/<topic>.md./livespec:critique) — file a critique-
style proposed-change (delegates to propose-change with
-critique reserve-suffix appended)./livespec:revise) — process all pending
proposed-changes with per-proposal accept/modify/reject
decisions, cut a new history/vNNN/ snapshot./livespec:doctor) — run static-invariant
checks across the main spec + each sub-spec tree. The
LLM-driven objective + subjective post-step phases are
not yet wired; only the static phase runs today./livespec:prune-history) — destructively
collapse old history/vNNN/ snapshots into a single
PRUNED_HISTORY.json marker. Requires explicit user
invocation; the LLM MUST NOT auto-activate this sub-command.Pointer to per-sub-command help. Mention that each
sub-command's wrapper supports -h / --help (e.g.,
bin/seed.py --help via Bash) for the CLI flag details.
No wrapper invocation; no post-step LLM-driven phase.
No wrapper exit codes — this skill is pure LLM narration. If the user asks about a capability that is not yet wired (e.g., the LLM-driven objective/subjective doctor phases), say so directly and point at the static phase as the currently-available behavior.