krypton-planning
Use when a user has a feature request, bugfix, refactor, migration, architecture change, or product goal and needs an implementation plan before coding. Use especially when wrong ownership, duplicate paths, stale contracts, weak evidence, or unclear cutover would make plausible agent work dangerous.
- License 许可
- Apache-2.0
- 最近更新
- 2026-11-14
- Stars
- ★ 1.6k
- Forks
- 491
01 前言 Front Matter
- name
- krypton-planning
- description
- Use when a user has a feature request, bugfix, refactor, migration, architecture change, or product goal and needs an implementation plan before coding. Use especially when wrong ownership, duplicate paths, stale contracts, weak evidence, or unclear cutover would make plausible agent work dangerous.
02 资源文件 2
- 参考文档agents/openai.yaml322 B
- 参考文档plan-reviewer-prompt.md1.2 KB
03 SKILL.md 正文
Krypton Planning
Krypton Planning turns a request into a goal package that names the outcome, truth owner, contract, cutover, and acceptance evidence before anyone writes code.
Core Rule
Do not treat a task list as a plan. A Krypton plan is ready only when it answers:
What outcome are we serving?
What current behavior is being replaced, redirected, deleted, or demoted?
Who owns the truth?
What contract crosses the boundary?
What is the smallest high-value slice?
What proves the result from the target person's perspective?
What kill criteria stop duplicate paths from living forever?
If ownership, cutover, contract, or evidence is fuzzy, stop and map before planning.
Workflow
- Write the outcome contract:
Plan title:
Intent:
Current behavior:
Expected outcome:
Target-perspective output:
Truth owner:
Contract boundary:
Cutover:
Displaced path:
Value density:
Evidence gate:
Acceptance evidence:
Evidence lane:
Kill criteria:
Non-goals:
Risk if wrong:
- Map the architecture slice before tasks:
Files to create:
Files to modify:
Files to avoid:
Source of truth:
Read path:
Write path:
Contract boundary:
Integration points:
Migration/cutover:
Displaced path:
Acceptance evidence gate:
For broad or unclear repositories, dispatch a read-only explorer if the harness supports agents. Ask one bounded question, such as "map the source of truth, read/write path, unsafe files, and evidence gate." The execution session should use this map instead of rediscovering the same slice.
- Save a goal package unless the user gives another location:
docs/goals/<goal-slug>/PLAN.md
docs/goals/<goal-slug>/GOAL.md
docs/goals/<goal-slug>/EVIDENCE.md
PLAN.md contains the full plan. GOAL.md is a short execution prompt that points to the plan instead of copying it. EVIDENCE.md starts empty and is filled during execution with the real proof captured from the target perspective.
- Start
PLAN.mdwith this header shape:
# [Outcome Title] Implementation Plan
**Intent:** ...
**Current Behavior:** ...
**Expected Outcome:** ...
**Target-Perspective Output:** ...
**Truth Owner:** ...
**Contract Boundary:** ...
**Cutover:** ...
**Displaced Path:** ...
**Value Density:** ...
**Evidence Gate:** ...
**Acceptance Evidence:** ...
**Evidence Lane:** ...
**Kill Criteria:** ...
**Architecture Slice:** ...
**Plan Review Gate:** Requires PRE review before execution.
-
Break work into small tasks. Each task names exact files, allowed scope, expected output, verification command, acceptance evidence, and whether it can run in parallel.
-
Run a PRE plan review when possible. Use
plan-reviewer-prompt.mdas the individual prompt template. Do not execute until blocker findings are fixed or explicitly accepted by the user.
GOAL.md Shape
Use this compact handoff. This is the /goal prompt or next-session prompt the operator should paste into Codex or Claude:
# Goal: [Outcome Title]
Use Krypton Execution to execute `docs/goals/<goal-slug>/PLAN.md`.
Core rules:
- Treat PLAN.md as the source plan.
- Preserve intent, ownership, contract, cutover, evidence, and kill criteria.
- Do not add a new dominant path without deleting, redirecting, demoting, or shimming the displaced path.
- Capture acceptance evidence from the target perspective and record it in EVIDENCE.md.
- Say "implemented but unproven" if that evidence cannot be captured.
EVIDENCE.md Shape
Create this file with the goal package and update it during execution:
# [Outcome Title] Evidence
## Acceptance Evidence
Record the real artifact that proves the outcome from the target perspective.
## Verification
Record the focused check that passed, including the command and relevant output.
## Review Notes
Record reviewer, maintainer, or verifier findings that changed the result.
04 同仓库更多 Skills
worktable-review-with-annotations
Create, inspect, reply to, update, or resolve situated Worktable annotations. Use when feedback belongs to a Doc section, block, text range, or HTML Doc, or when the user asks to address existing review feedback and coordinate a requested artifact edit.
worktable-find-and-synthesize-context
Find, assess, and synthesize context from Worktable Docs and Records. Use when the user asks what a workspace contains, wants workspace-backed research, needs current facts reconciled across artifacts, or requests a cited synthesis without changing workspace state.
worktable-create-or-update-docs
Create, edit, rename, or maintain narrative Worktable Docs. Use for plans, decisions, briefs, notes, research syntheses, and other durable prose when the user requests a new artifact or a safe update to an existing one.
worktable-create-or-manage-records
Organize repeated structured information as durable Worktable Records. Use for research datasets, inventories, directories, pipelines, trackers, and other collections whose items need identity, fields, filtering, links, validation, or individual updates.
worktable-create-or-manage-interactive-html
Create or improve durable visual and interactive artifacts in Worktable. Use for dashboards, trackers, calculators, mini apps, slide presentations, and other requests that benefit from a polished HTML Doc.
worktable-collaborate-threads
Read, answer, and hand off messages through durable Worktable threads with users or other authenticated participants. Use when the user asks to check addressed messages, reply to a participant, continue a thread, request another agent's help, wait briefly for a response, or watch for comments and messages sent to the agent.