| name | caspar-guide |
| description | Use when rendering the Next Steps footer at the end of any caspar command, suggesting which caspar command to run next, or answering "what command should I run?" / "how does the workflow work?". Do NOT use mid-task or for executing a phase — this only renders guidance and the footer. |
| user-invocable | false |
Caspar Guide
Renders the Next Steps footer after a command and answers which caspar command to run next. Routing + a fixed footer format only — not a tutorial.
Inputs
- Current phase + status (what command just ran, blocked or not).
- The command roster (below) — grounds every suggestion.
Working Set
Read just-in-time, never inline: the phase the caller is in and any blockers. Suggestions are drawn only from the roster below.
Outputs + DONE
A Next Steps footer appended to command output. DONE when:
- Footer states Phase · Status · Next (one concise recommendation) + up to 5 Options.
- Every option is either a real
/caspar: command from the roster or a non-slash manual action.
- Suggestions match the current stage; no command is invented.
Method / guardrails
Footer shape (plain text, no ASCII box):
Next Steps — Phase: {phase} | Status: {status}
Next: {1–2 line recommendation}
Options:
- /caspar:{command} — {why} # up to 5 total
- {manual action} — {why} # max 2 non-slash actions
Reply: {only if a textual reply is expected}
Rules (load-bearing):
- Status is exactly one of:
Active · Pending Input · Blocked · On Hold · Complete.
- Max 5 options, max 2 manual (non-slash) actions.
- Slash commands use the full
/caspar: prefix; manual actions carry no prefix.
- Never invent a command — suggest only roster entries. Only suggest commands that fit the current stage.
Command roster (S→P→E→C→R + standalone utilities):
| Command | When |
|---|
caspar-scope | start a feature — IN/OUT boundaries |
caspar-ux | UI-heavy — flows, components, states (after scope) |
caspar-plan | scope ready — research + route by complexity |
caspar-create_plan | complex work needing architecture first |
caspar-create_tasks | plan ready → concrete tasks |
caspar-plan_review | sanity-check a plan before task generation |
caspar-task_review | comprehensive task-artifact translation check |
caspar-execute | execute.md + tasks.json exist — multi-agent parallel build |
caspar-code_review | implementation complete - adversarial opposing-model review |
caspar-validate | verify implementation against scope |
caspar-create_test_guide | manual QA checklist |
caspar-clean | meta cleanup — prune, test, sweep |
caspar-prune | dead-code/artifact cleanup only |
caspar-test | risk-aware test coverage (P0–P3) |
caspar-rebase | rebase onto target, prep merge |
caspar-release | bump/push a Caspar version and refresh local Codex install |
caspar-fix | structured debugging for tough bugs |
Workflow shorthand: Scope → Plan → Execute → Clean → Rebase. prune, test, and sweep also run standalone; order is a default, not a requirement.
Handoff
The footer is the transition point — it tells the user/agent the next phase. Other caspar skills inline this same one-line Next-Steps convention at their close (no cross-skill reference to this skill).
Escalate-If
- The current phase has no sensible next command (e.g. blocked on external input) → set Status
Blocked/Pending Input and put the needed input in Reply, not a guessed command.
