status
Use for a quick, read-only snapshot of GOAL.json — goal, posture, focus, people, plan status, last eval, and recent log — without running a full strategy or eval cycle. Never writes to GOAL.json.
skyf0xx · 已有静态检查记录
安装与使用
运行命令前先阅读来源与依赖说明。复制命令不会自动安装或执行。
npx skills add skyf0xx/gambit --skill status技能说明
name: status description: Use for a quick, read-only snapshot of GOAL.json — goal, posture, focus, people, plan status, last eval, and recent log — without running a full strategy or eval cycle. Never writes to GOAL.json. display: plain-card
Skill: status
Trigger: You want a quick read on where things stand without a full strategy or eval cycle.
Purpose: Produce a lightweight snapshot from GOAL.json. Read-only — this skill never changes the plan or the focus, and never writes to GOAL.json.
Voice & Tone
Read-only reporting layer. Factual and terse — numbers and states, not interpretation. If asked "are we on track", pull the data and present what it shows; if it's ambiguous, say it's ambiguous and point to eval for a real audit.
Execution Sequence
Resolve GOAL.json per skills/_shared/RESOLVING.md and read it. If more than one goal
exists in the store, name which one this snapshot is for — a title alone is ambiguous
once the user is holding several.
Per AGENTS.md's "Opening the visualizer," open the live diagram view now if this is the
first Gambit skill run this session (nohup gambit visualize >/dev/null 2>&1 &, detached
and silent, mentioned once in passing) — don't relaunch it on a later call within the
same session. Output:
STATUS [date] — [goal title, if more than one goal exists in the store]
GOAL
[description, truncated if long]
Deadline: [date or none]
POSTURE
[current level/label, or omit this block entirely if the `posture` key is null]
FOCUS
[most recent non-null focus across log entries, or "none set"]
PEOPLE
[count confirmed / tentative, or omit this block entirely if the `people` key is empty]
PLAN
[one line per plan.linesOfOperation entry: "label — critical path — status", or "no plan yet"]
LAST EVAL
[from most recent `log` entry with `source: "eval"`, or "never run"]
RECENT LOG
[last 3-5 entries from the `log` key, one line each]
Close with a one-line menu — a snapshot with no route onward leaves the user holding data and no move:
Next: [strategy to reset focus | plan to sequence | brief for the plain-language read |
eval for a real audit | bmad-deep-recon to close a gap]
Recommend nothing. This skill reports; it doesn't steer — that's the distinction from
strategy. Just make the routes visible.
If resolution finds no goal: say so and point to onboard — it handles first-contact
intake properly, one question at a time, and also handles the case where several goals
exist and none is active.