How do I install the Dev Spec Kit Finish skill?

Run npx skillmds@latest add pratiyush/dev-spec-kit-finish in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Dev Spec Kit Finish safe to use?

This skill has not completed SkillMD's automated safety review yet. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Dev Spec Kit Finish?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Dev Spec Kit Finish free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Dev Spec Kit Finish?

pratiyush (@pratiyush) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Dev Spec Kit Finish by pratiyush — Agent Skill

Name: Dev Spec Kit Finish
Author: pratiyush

Dev Spec Kit Finish

The completion ritual — run when a feature's tasks are done and it's time to land the branch. Fresh-evidence entry gate, a fixed option menu, typed confirmation for destructive paths, provenance-checked cleanup, every step journaled. Use when the user says finish/land/merge/wrap up, or after dev-spec-kit pr is ready.

pratiyush 7259d8b 2.5 KB Updated Sep 17, 2026

File contents

dev-spec-kit-finish — landing a branch is a ritual, not a vibe

Entry is EARNED, the menu is FIXED, destruction is TYPED, and every step lands in the journal.

1. Entry gate — fresh evidence only

Run, in order, and SHOW the output (after any session break: dev-spec-kit resume FIRST):

dev-spec-kit graph build — MUST exit 0 (every proof green at the current tree). Stale/red → run dev-spec-kit drift, fix, and re-enter. Old green output does not count; the evidence must be fresh.
dev-spec-kit verify — MUST exit 0 (Build ALL + every configured kind green, journaled on the current tree). The PR gate refuses without it.
dev-spec-kit status — all tasks for this feature DONE.
dev-spec-kit pr --title "<title>" — body generated, gate verdict green. If any step fails: STOP. Report what blocked. Do not present the menu.

2. The menu — present exactly these 4 options, nothing open-ended

Merge locally — merge to the base branch, verify the suite on the merged result.
Push + PR — push the branch and open the PR with the generated body (dev-spec-kit pr --create).
Keep as-is — leave the branch for later; record why in the journal.
Discard — destructive. To proceed the human MUST type discard exactly. Anything else aborts.

Do not add commentary, recommendations, or a fifth option. Wait for the choice.

3. Execute + cleanup (option-conditional)

Merge: merge → re-run the suite on the merged result → only then clean up.
Push+PR / Keep: the worktree and branch SURVIVE (work may continue).
Discard (typed confirmation received): delete branch; remove the worktree ONLY if it lives under .worktrees/ or .claude/worktrees/ — provenance check: never clean up a workspace dev-spec-kit did not create.
dev-spec-kit approve <taskIds> --note "<the human's words>" if the human gate wasn't recorded yet.

4. Journal the landing

Every executed step is already audit-logged (cli.run); close with dev-spec-kit log -n 10 so the human sees the recorded trail: graph green → approval → PR/merge → cleanup. The ritual is complete only when the journal shows it.

---name: dev-spec-kit-finishdescription: The completion ritual — run when a feature's tasks are done and it's time to land the branch. Fresh-evidence entry gate, a fixed option menu, typed confirmation for destructive paths, provenance-checked cleanup, every step journaled. Use when the user says finish/land/merge/wrap up, or after dev-spec-kit pr is ready.---# dev-spec-kit-finish — landing a branch is a ritual, not a vibeEntry is EARNED, the menu is FIXED, destruction is TYPED, and every step lands in the journal.## 1. Entry gate — fresh evidence onlyRun, in order, and SHOW the output (after any session break: `dev-spec-kit resume` FIRST):1. `dev-spec-kit graph build` — MUST exit 0 (every proof green at the current tree). Stale/red → run `dev-spec-kit drift`, fix, and re-enter. Old green output does not count; the evidence must be fresh.2. `dev-spec-kit verify` — MUST exit 0 (Build ALL + every configured kind green, journaled on the current tree). The PR gate refuses without it.3. `dev-spec-kit status` — all tasks for this feature DONE.4. `dev-spec-kit pr --title "<title>"` — body generated, gate verdict green.If any step fails: STOP. Report what blocked. Do not present the menu.## 2. The menu — present exactly these 4 options, nothing open-ended1. **Merge locally** — merge to the base branch, verify the suite on the merged result.2. **Push + PR** — push the branch and open the PR with the generated body (`dev-spec-kit pr --create`).3. **Keep as-is** — leave the branch for later; record why in the journal.4. **Discard** — destructive. To proceed the human MUST type `discard` exactly. Anything else aborts.Do not add commentary, recommendations, or a fifth option. Wait for the choice.## 3. Execute + cleanup (option-conditional)- Merge: merge → re-run the suite on the merged result → only then clean up.- Push+PR / Keep: the worktree and branch SURVIVE (work may continue).- Discard (typed confirmation received): delete branch; remove the worktree ONLY if it lives under `.worktrees/` or `.claude/worktrees/` — provenance check: never clean up a workspace dev-spec-kit did not create.- `dev-spec-kit approve <taskIds> --note "<the human's words>"` if the human gate wasn't recorded yet.## 4. Journal the landingEvery executed step is already audit-logged (cli.run); close with `dev-spec-kit log -n 10` so the humansees the recorded trail: graph green → approval → PR/merge → cleanup. The ritual is complete onlywhen the journal shows it.

Frequently asked questions