Suede Recommend Next Action

Recommend one action and package it as a short runnable prompt. Inspect current state read-only; do not execute the recommended action unless the user separately authorizes execution. Keep the full operator contract hidden until the user asks to expand it.

Recommendation Workflow

1. Resolve the target and the user's actual done outcome from the current request, conversation, handoff, plan, repo, or live surface.
2. Check only the evidence needed to distinguish the next move. Prefer, in order: current terminal/repo/live state, current source documents, current plans or handoffs, then older memory.
3. Generate 2-4 candidate actions internally. Exclude work already verified as complete, adjacent cleanup, and actions outside the user's authorized scope.
4. Score each candidate from 0-2 on every criterion below. Recommend the highest total.

5. Break ties by preferring a required prerequisite, then current-evidence verification, then the more reversible action. If the top two remain within one point and target ambiguity would change the answer, run at most three additional read-only checks. If still tied, show both choices and state the single fact that decides between them.
6. Turn the recommendation into a 2-4 sentence quick prompt. Keep the scoring and full operator contract internal unless the user asks to compare choices, expand prompt, or make it granular.

Routing Rules

• If a repo or task already has its own plan, progress doc, issue tracker, or project board, do not create a second one. Treat its recorded next step as one candidate, verify it against current source, and recommend the winner — don't replace the existing tracker.
• If the user needs options explored before a commitment can be made, say so and offer to brainstorm instead of forcing a single recommendation.
• If missing evidence is the real blocker, make the smallest read-only check the recommended action and generate a prompt for that check.

Prompt Levels

Quick Prompt (Default)

Write 2-4 sentences that another capable agent can paste and run:

1. Name the selected skill or lane, exact target, and one concrete outcome.
2. Include the single current fact, scope boundary, or preservation rule most likely to change execution.
3. Name the observable done signal and its verification.
4. Add a stop condition only when a material blocker is plausible.

Do not dump field labels, the score breakdown, a long evidence list, or numbered implementation steps by default. After the prompt, add exactly:

Say "expand prompt" for the full operator version or "make it granular" for exact steps and commands.

Expanded Prompt (On Request)

When the user says expand prompt, render the full self-contained prompt below. Include only applicable fields and never leave placeholders such as TBD or <path>.

Use $<skill> to complete this task.

Target: <exact repo, worktree, route, URL, document, account, or surface>
Objective: <one concrete outcome>
Current verified state:
- <fresh fact and its source>
Authorized scope: <read-only, local edits, live mutation, or other exact boundary>
Preserve: <dirty work, copy, data, identities, or other invariants>
Source-truth order: <current sources in precedence order>

Required work:
1. <smallest complete sequence>

Done signal: <observable proof>
Verification:
- <command, readback, screenshot, URL, test, or response>
Stop and report if: <material ambiguity, unsafe mutation, failed gate, or missing authority>

Return: outcome, changed surfaces, verification evidence, caveats, and status.

Granular Prompt (On Request)

When the user says make it granular, render the expanded prompt and decompose Required work into atomic numbered steps. Include exact commands, absolute paths, mutation checkpoints, expected readbacks, and stop conditions wherever current evidence supports them. Do not invent missing commands or values; make resolving a missing decisive fact the first step.

For multi-lane work spanning more than one specialist, start expanded and granular prompts with $suede-agent-teams so the coordination lane owns the handoff. For a single-lane task, start with the specialist the recommendation names directly. Use exact absolute paths, URLs, handles, branch names, and verification commands when current evidence provides them.

Output Format

Recommended action: <one sentence>
Why now: <one evidence-backed sentence>

Quick prompt: <2-4 runnable sentences>

Say "expand prompt" for the full operator version or "make it granular" for exact steps and commands.

Show the route, score, evidence list, confidence, or alternatives only when the user asks for rationale or when the unresolved tie rule requires them. When the recommendation is an evidence-gathering step, state that in Why now without loading the expanded prompt.

Boundaries

• Do not mutate files, repos, deployments, accounts, messages, or live systems while recommending.
• Do not load the expanded or granular prompt by default.
• Do not repeat a broad audit when one current execution lane can be selected.
• Do not invent paths, URLs, skill availability, status, metrics, owners, or completion evidence.
• Do not recommend vague actions such as "keep working", "improve the app", or "do more research". Name a command, artifact, decision, edit, or verification result.
• Do not hide a blocker. If authority or a decisive fact is missing, make its resolution the next action.

Routing

• Need multi-lane coordination across specialists -> use suede-agent-teams.
• Need help picking which single skill fits a request -> read this pack's router (suede-workflow-skills) or ask directly.
• Need idea exploration before selecting a move -> brainstorm directly with the user instead of forcing a single recommendation.
• Need execution -> use the specialist named in the generated prompt.