菜单
Use when drafting, revising, reviewing, or promoting RFCs and design proposals, including Exo-managed RFCs, staged RFC lifecycle work, Rust/Ember-style RFC prose, and canonization or recommendation work for related RFCs.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Review the conversation for threads that were started but not followed — ideas mentioned in passing, decisions deferred implicitly, tensions named but not resolved. Use when taking stock, before saving, or whenever the user wants to know what's been left on the table.
Use when drafting, updating, or reviewing a GitHub pull request title or body, including PR creation flows that need reviewer-facing prose. Follow the repo PR template when present, preserve the user's or project's voice, and route validation/session details to the right surface.
End-of-day close: the user is stepping away and won't serve as a bridge. Full collaborative triage, working-style reflection, and anticipatory framing for the next day. Invoke with /session-close.
Start a new session cycle by reconciling Project Orientation, restoring repo context, and resuming the active bet or opening vision steering when selection genuinely changed.
End-of-cycle transition: interpret the session, triage dangling threads with the user, and prepare SESSION-TRAJECTORY.md for the next cycle. The user is still around and serves as a light bridge. Invoke with /session-rest.
Keeps SESSION-TRAJECTORY.md warm as work progresses and offers checkpoints at natural boundaries. Saving is not stopping.
| name | rfc-writing |
| description | Use when drafting, revising, reviewing, or promoting RFCs and design proposals, including Exo-managed RFCs, staged RFC lifecycle work, Rust/Ember-style RFC prose, and canonization or recommendation work for related RFCs. |
An RFC is durable public design reasoning. It makes a change understandable to users, implementers, maintainers, educators, and future readers, and it keeps the stage of the work visible as the design moves from idea to implemented reality.
Stances used: Load public-design-reasoning for the prose, collaborative-grounding and socratic-elicitation for the user's situated design intent, interpretive-synthesis for the coherent account, hypothesis-evaluating for stage readiness, and relational-continuity for precise and meaningful language that stays attached to the work underway.
The tension is public argument vs. implementation memory.
Recent context often contains branch state, debugging notes, design fragments, and half-formed decisions. An RFC turns that material into a public artifact. Write from the design the project is trying to make true, grounded in observed state and explicit evidence. Preserve tradeoffs, alternatives, and unresolved questions because they help future readers understand the decision, not because the template has empty boxes.
Template sections are reader contracts. Motivation helps readers understand why the change is worth considering. Detailed design makes the change buildable. Drawbacks make the tradeoff honest. Alternatives make the choice legible. Unresolved questions show where learning remains. Let each section do its reader-facing job in the current RFC.
When the repository is Exo-managed, treat Exo as the RFC source of truth. Exo owns RFC identity, lifecycle state, current stage, related goals and tasks, managed content, implementation evidence, and stage-transition readiness.
Discover the current RFC operations from the available Exo tool surface or the workspace's Exo documentation before reading or changing RFC state. Then use that managed surface to read lifecycle state, inspect the RFC artifact, update managed content, and request stage transitions. The plugin supplies the authoring stance and readiness judgment; Exo supplies the live operation contract.
Request a stage transition only when the user explicitly asks for it or confirms it after seeing the readiness evidence.
If Exo is unavailable, follow the repository's local RFC process and template. Inspect the local README, template, or prior accepted RFCs before choosing headings or format, and mark Exo-derived state as unavailable instead of reconstructing it from memory.
Match the prose to the lifecycle stage. Stage names vary by project; use the local process names and preserve these jobs.
Stage 0: Idea. Capture the live problem, affected audience, possible shape, and questions that determine whether the idea should receive more design attention. Keep the artifact light enough for exploration while making the problem concrete.
Stage 1: Proposal. Make the public case for doing the work. Explain the motivation, user and maintainer impact, design direction, drawbacks, alternatives, and the questions that must be resolved before acceptance.
Stage 2: Draft. Make the design implementation-ready. Specify the system contract, user-facing behavior, compatibility surface, migration path, testing shape, and clear criteria for reaching implemented reality.
Stage 3: Candidate. Reconcile the RFC with what was built. Read code, tests, docs, release notes, and Exo state before editing. The document should reflect implemented reality, including meaningful deviations from the draft and the evidence that supports readiness.
Stage 4: Stable. Align the RFC with shipped behavior and the living manual. Check that docs, tests, and current project behavior tell one coherent story, and that the RFC corpus remains mutually consistent.
Use canonization mode when a set of Stage 3 or Stage 4 RFCs together defines a recommended idiom, product feature, or project practice.
Synthesize the group as a corpus, not a pile of individual decisions. Identify the user-facing concept the RFCs now teach together, the implementation and tooling support that makes the idiom practical, the learning path that should appear in docs, and the migration or ecosystem state that makes broad use reasonable.
The output may be a recommendation section, a canonization note, a docs/manual update plan, or an RFC revision plan. In each case, make the recommendation rest on current evidence: shipped behavior, docs, tests, tooling, migration support, and consistency across the related RFCs.
Use a generate-then-evaluate loop so prose quality and readiness judgment do not collapse into each other.
repeat:
1. Observe the local process, stage, templates, existing RFCs, and evidence.
2. Ground the user's design intent and any situated audience concerns.
3. Draft or revise the RFC section by section according to each section's job.
4. Evaluate stage readiness against the evidence and local process.
5. Carry forward the concrete gaps, decisions, and next authoring work.
Ask the user for situated decisions: intent, priority, acceptable tradeoffs, audience, and whether to request a stage transition. Use tools for observable facts: RFC state, implementation status, tests, docs, git history, and local templates.
Exo currently models RFC Stages 0-4, with Stage 4 serving as the living canon. Treat Stage 5-style "Recommended" work as canonization mode in this skill: grouped, corpus-level work that may later deserve first-class Exo support.
Useful future Exo surfaces include recommendation groups, canonization checks, stage-specific authoring prompts, and manual/docs coherence checks. When the skill reveals one of those needs during real RFC work, name the Exo improvement as design pressure rather than mixing tool-design proposals into the RFC body.