Menu
Reminder to use rpce-cli
Installer avec Codex ou Claude Copiez ce prompt, collez-le dans Codex, Claude ou un autre assistant, puis laissez-le vérifier la page du skill et l'installer pour vous.
Basé sur la classification professionnelle SOC
Normalize a consequential or ambiguous request when missing context, boundaries, or proof could materially change the work. Explicit invocation previews the improved prompt.
Explore requirements and approaches through collaborative dialogue, then write a right-sized requirements document. Use when the user says "let's brainstorm", "what should we build", or "help me think through X", presents a vague or ambitious feature request, or seems unsure about scope or direction -- even without explicitly asking to brainstorm.
Find root causes and fix bugs. Use for errors, failing tests, issue repros, stack traces, "debug this", or "why is this failing".
Create structured plans for multi-step tasks once the goal is clear enough to plan. Use after ce-brainstorm or ce-grill, or directly for clear planning requests. If the request has branchy product/scope ambiguity, run ce-grill first. After markdown plans, document-review runs before handoff.
Review recent code changes for bugs, regressions, product fit, conventions, performance, security, and blast radius.
Execute implementation work with a compact product-contract loop. Use when the user asks to build, fix, implement, polish, or finish a scoped task. Favor reasoning, prior art, smallest correct changes, focused tests, and real-surface proof. Do not run autonomous PR, CI, ticket, or residual-work pipelines.
| name | rp-reminder-cli |
| description | Reminder to use rpce-cli |
| repoprompt_managed | true |
| repoprompt_skills_version | 61 |
| repoprompt_variant | cli |
Continue your current workflow using rpce-cli instead of built-in alternatives.
| Task | Use | Not |
|---|---|---|
| Search paths/content | search | grep, find, Glob |
| Read file (whole or sliced) | read | cat, head, Read |
| Directory tree | tree | ls, find |
| Signatures / overview | structure | reading whole files |
| Edit file | edit | sed, Edit |
| Create / delete / move | file | touch, rm, mv, Write |
| Git status / diff / log / blame | git | shelling out for analysis |
| Tool | Use for |
|---|---|
manage_selection | Curate the file set used by chat, builder, and exports. Refresh before each planning call. Modes: full, slices, codemap_only. |
workspace_context | Snapshot current prompt + selection + token budget; also exports. |
prompt | Read/set the shared prompt; list or select copy presets. |
context_builder | Heavy discovery sub-agent — describe the task, it curates files + rewrites the prompt. response_type: clarify / plan / question / review. Pass export_response:true to hand the result to a child agent. |
chat (ask_oracle) | Chat-mode reasoning over the current selection. Continue existing chats (new_chat:false) rather than opening new ones. Modes: chat / plan / review. |
oracle_chat_log | Recover recent Oracle messages after compaction. |
ask_user | Ask the user when ambiguity is load-bearing — don't guess at requirements. |
agent_run / agent_manageDispatch a sub-agent when a side investigation or delegated chunk of work would otherwise flood this session's context.
Role labels (pass as model_id on agent_run op=start):
| Role | Use for |
|---|---|
explore | Fast read-only probes — git archaeology, "where is X wired?", narrow lookups, web/doc search. One question per probe. |
engineer | Balanced implementation work delegated to a child agent. |
pair | Multi-step reasoning with back-and-forth — lead investigator or main implementer of a decomposed item. |
design | Architecture / review / extended analysis — primary deliverable is a markdown report under docs/reviews/, docs/designs/, or docs/analysis/. Expect the report path in the summary. |
Key agent_run ops: start (creates a new session/tab — never pass session_id here), wait / poll (accept session_id or session_ids array), steer (continue an existing session), respond (answer a pending interaction_id), cancel.
Key agent_manage ops: list_agents (discover roles + compound model_ids), list_sessions, get_log, cleanup_sessions (delete finished MCP-started sessions).
Fan-out pattern: call agent_run op=start with detach:true for each probe, then agent_run op=wait session_ids=[…] to block on the batch. Always follow a detach with a wait — don't leave probes unattended.
Export handoff: when context_builder or ask_oracle returns oracle_export_path, include that path inside the child agent's next message so it reads the export with read_file.
# Search · Read · Edit · File ops
rpce-cli -w <window_id> -e 'search "keyword"'
rpce-cli -w <window_id> -e 'read Root/file.swift --start-line 50 --limit 30'
rpce-cli -w <window_id> -e 'call apply_edits {"path":"Root/file.swift","search":"old","replace":"new"}'
rpce-cli -w <window_id> -e 'file create Root/new.swift "content..."'
# Selection · Builder · Oracle
rpce-cli -w <window_id> -e 'select add Root/path/file.swift'
rpce-cli -w <window_id> -e 'builder "<task>" --response-type plan'
rpce-cli -w <window_id> -e 'chat "..." --mode plan'
# Delegate · Fan-out · Steer · Cleanup
rpce-cli -w <window_id> -e 'agent_run op=start model_id=explore session_name="Probe: X" message="<question>" detach=true'
rpce-cli -w <window_id> -e 'agent_run op=wait session_ids=["<uuid1>","<uuid2>"] timeout=60'
rpce-cli -w <window_id> -e 'agent_run op=steer session_id="<uuid>" message="now do Y" wait=true'
rpce-cli -w <window_id> -e 'agent_manage op=cleanup_sessions session_ids=["<uuid>"]'
Continue with your task using these tools.