sdd-lifecycle
Deliver non-trivial Personal Assistant work with explicit plan-only, sdd-only, implement-approved-spec, and full-lifecycle modes.
Skill by edonyzpc
Install
Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.
npx agentmarket install skill/edonyzpc-sdd-lifecycle
Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.
Get the weekly diff
The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.
About
Use cases: design
Reviews
Write a reviewNo reviews yet. Installed it? Reviews come from verified installers only. How reviews work
Permissions
What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.
- Tools
- No tools found
- Secrets
- None asked for
- Install scripts
- None found
Network hosts it talks to 1
codeload.github.com· found in install command
Files it touches
~/.claude/skills/sdd-lifecycle
What's inside
- Skill name
sdd-lifecycle- Description
- Deliver non-trivial Personal Assistant work with explicit plan-only, sdd-only, implement-approved-spec, and full-lifecycle modes. Use when routed by pa-docs-lifecycle-manager or explicitly asked to plan, design, implement approved scope, continue an active track, or close it. Create Plan and SDD only when complexity requires them, stop plan-and-implement after validated implementation, and keep closeout, commit, push, tag, publish, and release separately authorized.
- License
- AGPL-3.0
- Instructions
- 126 lines in SKILL.md · view on GitHub
Files (2)
- SKILL.md
- openai.yaml · agents/
SKILL.md (first lines)
# SDD Lifecycle ## Boundaries Choose the mode covering the full authorized request; complete its stop point without pausing at intermediate modes. A still-applicable `review-only`, `analysis-only`, `read-only`, `no-file-changes`, “只分析”, or “不要改文件” request means **zero writes**. | Mode | Allowed work | Stop point | | --- | --- | --- | | `plan-only` | Scope approved work; create a Plan only when phased/risky delivery needs one | Before implementation design/code | | `sdd-only` | Write or review source-verified implementation design | Before runtime edits | | `implement-approved-spec` | Create justified prerequisites, implement, test, review, fix, and validate | Validated implementation; no closeout/commit | | `full-lifecycle` | Delivery plus explicitly requested closeout | Contracts reconciled and process artifacts disposed | “Plan and implement” is `implement-approved-spec`; it never implies closeout, archive, or commit. If target resolution remains ambiguous after explicit ID/slug, current-conversation package, and the only Active Package, ask one target question and perform zero writes. ## Read Set Read `AGENTS.md`, the owning Product Spec or Governance Contract, the matching Tracker, and nearby architecture/code/tests. Read North Star for product work. Read Plan/SDD only when they exist or the change justifies creating them. Treat Archive as historical evidence, not approval. ## Artifacts - Use exactly one authority lane: Product Decision/Spec for product behavior; Governance Contract for repo-only docs/checker/CI/release/Agent rules. - Baseline Active Package: `README.md` + `tracker.md`. - Add `plan.md` for phased delivery, material dependencies, risk, rollback, or cross-session execution. - Add `sdd.md` for multi-module design, shared infrastructure, behavior/data/ privacy/lifecycle changes, compatibility, migration, or non-trivial UI state. - Tracker is the only execution status and validation log. Active Registry and Feature Home contain links, not status mirrors. - Enforce `1 Now + 1 Next`: at most one `Implementing | Validating | Blocked` package plus one `Planned` package. At `Validated`, continue already-authorized closeout; ask once only if closeout authority is missing. - Tracker `Current Snapshot` is the cross-session handoff. Never create standalone `handoff*.md` or `closeout.md` in an Active Package. - Follow `docs/development/documentation-workflow.md` for exact lifecycle and retention rules; do not duplicate its templates inside this skill. ## Plan And Design Before implementation: 1. Verify approved product scope or explicit governance authority from current contracts and the conversation; do not ask again for an already-made choice. 2. Search the actual dependency surface with `rg`. For PA Agent/runtime/command work, apply the [Command Architecture Contract](../../../docs/architecture/pa-agent-architecture-plan.md#command-architecture-contract). Map this task's material decisions, admission, state transitions and effects to owners and factual bases in the existing SDD/task record; do not copy a second definition or infer the intended architecture from local command code. 3. Create/update Feature Home and Tracker; register the Feature Home. 4. Create a Plan only when its delivery/risk content would not fit concisely in Tracker. 5. Create an SDD only when the design criteria above apply. Verify every named method, type, setting, command, locale key, and CSS class against source. 6. Record compatibility, rollback, and a requirement-to-test mapping where relevant. Close or explicitly defer P0/P1/P2 design findings. Missing Plan/SDD in `implement-approved-spec` is not a separate user approval gate: create only what the implementation genuinely requires. Stop if product scope or risk acceptance is unresolved. ## Implement And Validate For each behavior slice: ```text implement -> focused validation -> review -> fix -> verify ``` Keep the Tracker current and use existing module boundaries. Before a slice, map its REQ/AC or risk to the change, minimum sufficient evidence, pass condition, and rerun/expansion trigger. Follow **Validation Planning And Reuse** and **Test Failure Diagnosis** in `AGENTS.md`; add tests for uncovered behavior or regression risk, and select the correct source/tooling/artifact group. Satisfy the Local Validation Gate at the justified scope, reusing covered checks only with verified inputs/results. Use PA review/follow-up skills for implementation review; fix-driven verification covers the affected risk. Accept responsibility conformance before implementation/green tests. Separate deterministic harness, actual-model semantics/recovery and real-app evidence according to the affected boundary; fixtures cannot prove model generalization. Retain required broad and app/device gates at each phase exit. For app proof, use `make deploy` or the eligible current-build reuse target under `AGENTS.md` Local Deployment, then observe the affected behavior. Do not claim app or iOS validation without observed evidence; docs/skills-only work uses the docs gate. ## Closeout Enter only in `full-lifecycle` or after explicit close/archive intent. 1. Reconcile actual behavior, current Product/Architecture/Governance contract, Tracker, and release state. 2. Record final checks, residual risk, and artifact disposition in Tracker; move unresolved work to Backlog. 3. Absorb durable outcomes into current contracts/tests. 4. Delete Feature Home, Tracker, Plan, SDD, handoff, and round-by-round logs after absorption unless one contains unique evidence. 5. Retain only a compact closeout or evidence document when a current authority needs the historical rationale. Place it in `docs/archive/<year>/` and link it from that authority; do not preserve a complete package by default. 6. Update the link-only Active Registry and run docs validation. If a selected archive path exists, fail closed before mutating source state; do not overwrite, merge, auto-suffix, or partially move it. ## Git Boundary And Output Do not stage, commit, push, tag, publish, or release without explicit authority. For a requested local commit, inspect status and focused diffs, stage only the
From edonyzpc/personal-assistant, AGPL-3.0.
More from edonyzpc/personal-assistant
- 0.789score
obsidian-test-vault-smoke
edonyzpc
Validate Personal Assistant runtime and UI changes in the repo-local Obsidian test vault.
SkillAFreshness: A★ 149Claude Code +9 - 0.785score
obsidian-dataview
edonyzpc
Dataview plugin query syntax, inline expressions, DataviewJS API, and common vault analysis patterns.
SkillAFreshness: A★ 149Claude Code +9 - 0.783score
stable-release
edonyzpc
Execute the stable release workflow for the Personal Assistant Obsidian plugin.
SkillAFreshness: A★ 149Claude Code +9 - 0.783score
personal-assistant-review-followup
edonyzpc
Triage, confirm, fix, and validate code review findings in the personal-assistant repository after an agent team or human review.
SkillAFreshness: A★ 149Claude Code +9 - 0.783score
personal-assistant-review
edonyzpc
Review uncommitted or PR diffs in the personal-assistant Obsidian plugin with project-specific risk lanes, second-layer future-risk checks, severity…
SkillAFreshness: A★ 149Claude Code +9 - 0.783score
ui-ux-design-audit
edonyzpc
Audit multi-surface UI/UX in the personal-assistant Obsidian plugin with the current PA framework, real interface evidence, structured scoring, adversarial…
SkillAFreshness: A★ 149Claude Code +9 - 0.780score
obsidian-community-check
edonyzpc
Trigger and inspect Obsidian Community checks for personal-assistant.
SkillAFreshness: A★ 149Claude Code +9 - 0.780score
obsidian-ios-real-device-smoke
edonyzpc
Validate Personal Assistant on an iPhone using the iCloud test vault and Safari Web Inspector.
SkillAFreshness: A★ 149Claude Code +9 - 0.780score
pa-docs-lifecycle-manager
edonyzpc
Maintain PA task records and documentation lifecycle.
SkillAFreshness: A★ 149Claude Code +9 - 0.779score
pa-brat-beta-release
edonyzpc
Manage Personal Assistant BRAT beta prerelease workflow.
SkillAFreshness: A★ 149Claude Code +9 - 0.779score
obsidian-templater
edonyzpc
Templater plugin syntax, template commands, dynamic commands, user scripts, and startup templates.
SkillAFreshness: A★ 149Claude Code +9 - 0.777score
pm-needs-analysis
edonyzpc
分析 PA 功能请求、用户需求与产品匹配度,依据已有信息给出有证据的判断和建议;用户选择访谈时逐层引导,未定产品决策交由用户。
SkillAFreshness: A★ 149Claude Code +9
Signals
- Stars
- 149
- Forks
- 8
- Open issues
- 10
- Last commit (this folder)
- Oct 4, 2026 (24 days ago)
- Last release
- Oct 24, 2026 (4 days ago)
From GitHub (fetched Oct 24, 2026).
Details
- License
- AGPL-3.0
- Repository
- github.com/edonyzpc/personal-assistant
- First seen
- Oct 24, 2026
- Last checked
- Oct 24, 2026
- Source
- GitHubgithub-graphql
More from edonyzpc
- 0.789score
obsidian-test-vault-smoke
edonyzpc
Validate Personal Assistant runtime and UI changes in the repo-local Obsidian test vault.
SkillAFreshness: A★ 149Claude Code +9 - 0.785score
obsidian-dataview
edonyzpc
Dataview plugin query syntax, inline expressions, DataviewJS API, and common vault analysis patterns.
SkillAFreshness: A★ 149Claude Code +9 - 0.783score
personal-assistant-review
edonyzpc
Review uncommitted or PR diffs in the personal-assistant Obsidian plugin with project-specific risk lanes, second-layer future-risk checks, severity…
SkillAFreshness: A★ 149Claude Code +9 - 0.783score
ui-ux-design-audit
edonyzpc
Audit multi-surface UI/UX in the personal-assistant Obsidian plugin with the current PA framework, real interface evidence, structured scoring, adversarial…
SkillAFreshness: A★ 149Claude Code +9 - 0.783score
personal-assistant-review-followup
edonyzpc
Triage, confirm, fix, and validate code review findings in the personal-assistant repository after an agent team or human review.
SkillAFreshness: A★ 149Claude Code +9 - 0.783score
stable-release
edonyzpc
Execute the stable release workflow for the Personal Assistant Obsidian plugin.
SkillAFreshness: A★ 149Claude Code +9
Cite this page
sdd-lifecycle skill. Agent Mart. https://agentmart.dev/skills/edonyzpc-sdd-lifecycle (last updated 2026-10-24)