openspec-propose
Propose a new change with all artifacts generated in one step. Use when the user wants to quickly describe what they want to build and get a complete proposal with design, specs, and tasks ready for implementation.
- License 许可
- MIT
- 所属仓库
- cp-yu/edera
- 最近更新
- 2026-06-23
- Stars
- ★ 2
- Forks
- 0
01 前言 Front Matter
- name
- openspec-propose
- license
- MIT
- metadata
- { "author": "openspec", "version": "1.0", "generatedBy": "1.4.1-cpyu.1" }
- description
- Propose a new change with all artifacts generated in one step. Use when the user wants to quickly describe what they want to build and get a complete proposal with design, specs, and tasks ready for implementation.
- compatibility
- Requires openspec CLI.
03 SKILL.md 正文
Propose a new change and generate all artifacts needed for implementation.
Flow
- Input must identify a kebab-case change name or enough description to derive one. If unclear, ask what to build or fix.
- Apply smart routing before creating files: inspect the current conversation for an explore-generated
Design Summary; if no summary exists, respectpropose.smartRouting: falseandpropose.requireExplore: false, otherwise score the user's input across 5 dimensions. Detect multi-subsystem scope. Outcomes include: "Design Summary found: proceed and show that Design Summary is being used", "Input is sufficiently detailed. Skipping explore; generating artifacts directly.", and "This request spans multiple independent subsystems. Consider running$openspec-exploreto decompose it first.". Show input length, detail score, multi-subsystem result, and final decision. - Run
openspec new change "<name>", thenopenspec status --change "<name>" --jsonto readapplyRequires, artifact order, dependencies, and schema. - Load shared OPSX context before artifact generation.
Before reading other context files, check whether
openspec/project.opsx.yamlexists.
- If it exists, read it first for domains → capabilities structure
- Read the
project:block for project intent and scope - Treat it as navigation context, not as a replacement for change artifacts
- Before specs, run
openspec list --specs --json; compare proposed capabilities to each spec'scapabilitiesstring array. Specs without frontmatter returncapabilities: []. Reuse or modify existing coverage instead of duplicating specs. - Use CLI-backed OPSX navigation after shared context.
After reading shared
project.opsx.yamlcontext, use OpenSpec CLI query surfaces for node details.
- Run
openspec list --specs --jsonto get specs and theircapabilitiesstring arrays; specs without frontmatter returncapabilities: []. - For known or affected OPSX node IDs, run
openspec opsx query <node-id...> --jsonto get node details, relations and code-map refs in one batch; add--depth 2when broader related context is needed. - Treat CLI output as navigation context, not as a replacement for change artifacts.
- For each ready artifact, run
openspec instructions <artifact-id> --change "<name>" --json; readconfigProjection(especiallyconfigProjection.normalized.proseLanguageandconfigProjection.prompt.fragments), dependencies,template,instruction, andoutputPath; follow the template exactly and do not copycontext,rules, orconfigProjectioninto artifact files. When creatingspecs, apply the returnedSpec content boundary: route non-behavior content to design/tasks/proposal/opsx-delta instead of requirements. - Continue until all
applyRequiresartifacts aredone. If an artifact is unclear, ask one focused question and continue. - After spec-driven specs are complete, generate
opsx-delta.yamlfromopenspec instructions opsx-delta --change "<name>" --json; useschema_version: 1,ADDED:,MODIFIED:, andREMOVED:YAML keys and query existing nodes when needed. - Run warning-only post-propose validation: This validation is warning-only. Do NOT turn
$openspec-proposeinto a blocking gate. Preferopenspec validate "<name>" --type change --json; align withValidator.validateChangeDeltaSpecs(), SHALL/MUST requirement text, required#### Scenario:blocks,Validator.validateOpsxDelta(),applyOpsxDelta(), referential integrity, and code-map integrity. Do NOT runopenspec sync; report when validation skips this check. For structure checks, readopenspec instructions proposal --change "<name>" --json,openspec instructions design --change "<name>" --json, andopenspec instructions tasks --change "<name>" --json; usevalidateTaskStructure, support Actions and coarse### Task N:with Goal, Files, Requirements, Checks, Covers:, Verifies:, change-localVerifies:spec paths, Requirement/Scenario references, Command:, Evidence:, and Expect:. Do NOT invent semantic lint rules beyond the current templates. Do NOT judge whether a check is semantically sufficient. If warnings appear, do exactly one repair pass, re-check once, and summarize remaining warnings. - Finish with
openspec status --change "<name>"and report artifacts created plus readiness for$openspec-apply-change.
Artifact Contract
Document Language Contract:
- Treat
openspec/config.yamlas the compact source of truth, but consume its compiled prompt projection rather than reinterpreting raw keys ad hoc - If the compiled projection includes
proseLanguage, apply it to natural-language prose you write or revise in the artifact body - Natural-language prose includes task titles, check names, Requirement titles, Scenario titles, bullet descriptions, Expect/Evidence descriptions, rationale, goals, risks, and summaries
- Follow the existing template structure exactly; do not invent a different layout because the prose language changes
- Keep template headings, normative keywords, BDD keywords, IDs, schema keys, relation types, file paths, commands, and code identifiers in their canonical form
- Preserve exact existing Requirement titles required for MODIFIED matching
- English project terminology may remain embedded in prose, but ordinary English sentences and titles still follow
proseLanguage - If no
proseLanguageprojection is present, keep the default writing behavior for prose
Keep generated tasks coarse: ### Task N:, Goal, Files, Requirements, and nested Checks. Keep each task to at most 5 Requirements. Preserve template structure, canonical headings, IDs, schema keys, paths, commands, BDD keywords, and code identifiers.
04 同仓库更多 Skills
openspec-snack
Quick code-first sync: generate or update proposal + specs + simplified design from git diff when the code is already written. Use after iterative coding to back-fill specs and OPSX delta without redoing propose→apply. Does not generate tasks.md.
openspec-reviewer
Internal clean-context Phase 1 verification reviewer. Judges implementation completeness, correctness, coherence, and cleanliness by reading files from changeName, changeDir, and projectRoot. Never accesses conversation history.
openspec-propose
Propose a new change with all artifacts generated in one step. Use when the user wants to quickly describe what they want to build and get a complete proposal with design, specs, and tasks ready for implementation.
openspec-optimizer
Internal clean-context Phase 2 optimization proposer. Analyzes implementation files and outputs behavior-preserving Search/Replace blocks. Never modifies files directly. Reads failedDirections to avoid repeating broken strategies.
openspec-impact-sweeper
Generate a lightweight OPSX-grounded JSON impact report for one project concept. Use from explore before scope or proposal readiness claims.
openspec-explore
Enter explore mode - a thinking partner for exploring ideas, investigating problems, and clarifying requirements. Use when the user wants to think through something before or during a change.