/suede-recommend-next-action · Agent orchestration & workflows; the pack's decision layer
One next move, and the prompt you can paste to run it.
It reads the repo, the terminal, the plan, or the handoff, read-only, and returns a single move with a short self-contained prompt. Behind that answer it scored two to four candidates on goal fit, unblocking, evidence, urgency, and leverage, so the recommendation can be argued with on the record. Ask for the prompt expanded or made granular and it does that too.
"what should I do next"->/suede-recommend-next-actionAsk in plain words. The router reads the request and loads this lane; MCP agents find it with list_suede_skills.
claude code · sample session
$ /suede-recommend-next-action what's next3 candidates scored · goal fit, urgency, leverageONE MOVE · RUNNABLE PROMPT READYSay "expand prompt" for the full operator version
Every candidate is scored 0-2 on five criteria; the highest total wins. Ties break toward the required prerequisite, then the action confirmable by current evidence, then the more reversible choice.
Criterion
2 points
1 point
0 points
Goal alignment
Directly produces the done signal
Required prerequisite
Merely adjacent
Unblocking
Unlocks a core path or 2+ steps
Unlocks one step
Unlocks nothing known
Evidence
Confirmed by current source
Confirmable with one check
Depends on an assumption
Urgency
Active failure, deadline, or gate
Needed for the milestone
No current pressure
Leverage
One session, prevents rework
Bounded, moderate payoff
Unscoped or low-payoff
What it covers
Read-only inspection of current terminal, repo, plan, or handoff state; never mutates files, deployments, or accounts while recommending.
A default quick prompt: 2-4 runnable sentences naming the skill, target, outcome, one decisive fact, and a done signal.
An expanded prompt on request ("expand prompt"): full operator contract with target, objective, verified state, authorized scope, and verification steps.
A granular prompt on request ("make it granular"): the expanded prompt decomposed into atomic numbered steps with exact commands.
Refuses vague recommendations like "keep working" or "improve the app"; names a command, artifact, decision, edit, or verification result.
Defers to a repo's existing plan, progress doc, or issue tracker instead of creating a second one.
When to use it
The user asks "what's next," "what should I do next," or "recommend the next move," especially after a review, audit, plan, or stalled task.
The user asks to "give me the prompt," "expand prompt," or "make it granular."
Choosing between two or more plausible next steps with limited context to spare.
NOT for executing the recommended action without separate authorization.
NOT for coordinating a multi-lane build across specialists; use suede-agent-teams instead.
Best prompt
Use $suede-recommend-next-action to give me one next move and a short runnable prompt. Expand it only if I ask.
Related skills
Pairs with suede-agent-teams for multi-lane coordination once a recommendation spans more than one specialist, and with suede-workflow-skills, the umbrella router that loads whichever lane this skill names.
Install this skill
It ships inside the Suede Creator Skills pack, so one install brings the whole pack and this skill with it. Claude Code takes two commands the first time and one after that; Codex installs the same pack natively.