finish-work
Use before you say that work is complete or done, before you commit or push, when you decide which tests, checks, or gates to run, when a gate or a test suite…
Skill by futex-ai
Chalk Labs
Need help deploying finish-work?
Chalk Labs sets up agents like this inside real teams: connected to your tools, tested on your data, with a human in the loop where it matters.
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/futex-ai-finish-work
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: testing
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/finish-work
What's inside
- Skill name
finish-work- Description
- Use before you say that work is complete or done, before you commit or push, when you decide which tests, checks, or gates to run, when a gate or a test suite fails or a run was partial, when you add a feature that needs a smoke test, and when the change that you finish is documentation-only or plan-only.
- License
- MIT
- Instructions
- 93 lines in SKILL.md · view on GitHub
Files (1)
- SKILL.md
SKILL.md (first lines)
# Finish work Work is complete only when steps 1 to 7 below are done, in order. Step 8 comes after the merge. The commands for each step come from the repository's `AGENTS.md` under `## Commands`. If a step cannot run, say so, name the blocker, and list the checks that did run. ## 1. During development - Run only the tests that cover the change. Require a 100% pass rate for the tests that run. - Run each new asynchronous or browser test 10 times before the commit, as the `testing` skill says. - Run the early repository check (format, lint, file length, exports) early, before long test runs. - Leave complete unit and browser suite runs to the complete gate. - When you add a feature, run a smoke test: start the server, run the command, open the page. Do this before the gate. ## 2. The complete gate - Run the complete gate once before you say work is complete. - A local run usually stops at the first failed suite. After a failure, fix it. Rerun only the failing tests, or the failed suite. Then rerun the complete gate. - Any selected test run, filtered run, or single suite is partial verification. Only the complete gate proves the work. - If a test that the diff does not touch fails and then passes on a rerun, it is an unrelated flaky test. Do not fix it in this branch. Report it under the rules in the `deterministic-tests` skill. - A flaky test that the diff adds or changes may be fixed without asking only under the flaky-test rule in the `deterministic-tests` skill. - Documentation-only or plan-only changes, including initial plan creation, do not require the gate. Validate the changed Markdown and review the diff instead. For a plan with a new decision, run the options check over it, as the `plans` skill says. ## 3. Keep the spec aligned - Docs, protocol docs, mockups, and every `README.md` are part of the spec. Update them in the same change. A protocol doc changes only when a requirement or a decision changes. A README changes when an entry point or the public behaviour changes. See the `documentation` skill. - Tick completed plan TODOs and add discovered TODOs. See the `plans` skill. - When the change removes or renames anything, search `docs/`, `plans/`, and every `README.md` for its name and update live references. See the `git-commit` skill. ## 4. Commit and push - Commit with the `git-commit` skill: run its deletion audit before and after the commit, run `git add -A` so that newly created files are tracked, and write a Conventional Commits message. Then push the branch. - If the branch needs a merge or rebase first, follow the `git-merge` skill. - A new plan gets no pull request in this step. Step 6 says when it opens. See the `plans` skill. ## 5. Review - After the push, review the complete local diff against `origin/main` with the `implementation-review` skill. Review a branch that targets a primary branch against that branch instead. The review itself is read-only. Its findings stay in the diff unless the user explicitly asked for a wider review. - Record each round in the findings file of the branch, as the `implementation-review` skill says. - Before each fix round, run the finding checks of the `implementation-review` skill: one check for each new finding. Each check confirms its finding, finds its links, and checks its options. A finding that is not real or not in the scope is removed. The check corrects a wrong severity or category in place. - Apply the review-fix rule. Make the final call on the outcome of each confirmed finding, with the merged options. Fix the findings with `Action: fix now`, with the option that each names. Record a follow-up item for each other finding and for each wider fix that you split off, as the `plans` skill says. Run the checks. Commit and push. Re-review once. Then stop the fix rounds and go on to step 6. ## 6. Pull request - The implementation is complete when steps 1 to 5 are done. When the change has a plan, the implementation is complete only when the plan's other milestone TODOs are also ticked. A new plan has unticked milestone TODOs, so it is not complete. Open findings that wait for the user do not make the implementation incomplete. - When the change has a plan, read "Deleting the plan" in the `plans` skill before you open the pull request. It says when the plan goes, when it stays, and the steps. - Open the pull request when the implementation is complete, or when the user asks for it. Before that, report the branch. Write the title and the description with the `git-commit` skill. The description lists each follow-up item that the change adds, so that the user can bring one back into the branch before the merge. ## 7. Report Write the final message in Simplified Technical English, with the layout and labels in the `decisions-and-findings` skill. Include: - Each item that still needs the user's answer, also an item from an earlier message. Name it in the first sentence. Give it in full, as "Open items" in the `decisions-and-findings` skill says. - What changed, the commit, and the branch or the pull request. - The checks that ran and their results. Name any check that was skipped or could not run, and why. - The test cost of the change: each test that it adds or removes, with its run time, as the `testing` skill says. - The fixed review findings, the deferred findings and wider fixes with their follow-up items, and the findings that need a decision, each with a recommendation. Keep them in three lists. - The number of open follow-up items in the repository. - The follow-up items of the blocked wider fixes, one line per class. - Problems outside the review scope, for a separate branch. An unrelated flaky test gives its name, failure text, rerun outcome, and suspected source. - Open plan TODOs. ## 8. After the merge
From futex-ai/skills, MIT.
More from futex-ai/skills
- 0.592score
git-commit
futex-ai
Use before you commit and directly after, when you write a commit message or a pull request title and description, when you squash, when the diff against…
SkillAFreshness: A★ 1Claude Code +9 - 0.592score
writing-skills
futex-ai
Use when you create or edit a skill, write or change a skill description, or when an agent skipped a rule, did not load a skill, or loaded the wrong one.
SkillAFreshness: A★ 1Claude Code +9 - 0.592score
deterministic-tests
futex-ai
Use when a test fails and then passes on a rerun, when a test is flaky or timing-dependent, when a test would assert elapsed time, a timeout, or performance…
SkillAFreshness: A★ 1Claude Code +9 - 0.592score
shared-types
futex-ai
Use when a Rust backend and a TypeScript or JavaScript frontend exchange data, before you add or change a request, response, query, event, message, or error…
SkillAFreshness: A★ 1Claude Code +9 - 0.592score
rust
futex-ai
Use before you write or change Rust code, add a crate or a dependency to a Cargo workspace, write or move a Rust test, define or map an error type, or add a…
SkillAFreshness: A★ 1Claude Code +9 - 0.592score
implementation-review
futex-ai
Use after you push a branch and the work needs a review, before you run a review or apply its findings, when a plan's review TODO is reached, when the user…
SkillAFreshness: A★ 1Claude Code +9 - 0.589score
git-merge
futex-ai
Use before you merge or rebase a branch, when origin/main has moved ahead of your branch, when you resolve merge conflicts, directly after you commit a merge…
SkillAFreshness: A★ 1Claude Code +9 - 0.587score
product-ui
futex-ai
Use before you implement or change a screen, a component, a view, or a mockup, when you write or edit user-facing text such as a status, helper, label, or…
SkillAFreshness: A★ 1Claude Code +9 - 0.587score
decisions-and-findings
futex-ai
Use before you send a message that asks the user for a decision, presents review findings or review items, or gives a final report, when the user asks for…
SkillAFreshness: A★ 1Claude Code +9 - 0.587score
documentation
futex-ai
Use before you edit a README.md, a crate README, anything under docs/, an architecture doc, or a protocol doc, when a behaviour change could leave a doc…
SkillAFreshness: A★ 1Claude Code +9 - 0.587score
typescript
futex-ai
Use before you write or change TypeScript or JavaScript, add an npm dependency, set up a new package, or write a frontend file, a Node script, or a test file…
SkillAFreshness: A★ 1Claude Code +9 - 0.587score
databases
futex-ai
Use before you add, edit, or name a database migration or a seed migration, when you touch a schema, Diesel code, a migration lock, or a crate named…
SkillAFreshness: A★ 1Claude Code +9
Signals
- Stars
- 1
- Forks
- 0
- Open issues
- 0
- Last commit (this folder)
- Oct 20, 2026 (11 days ago)
From GitHub (fetched Oct 24, 2026).
Details
- License
- MIT
- Repository
- github.com/futex-ai/skills
- First seen
- Oct 24, 2026
- Last checked
- Oct 24, 2026
- Source
- GitHubgithub-graphql
More from futex-ai
- 0.592score
shared-types
futex-ai
Use when a Rust backend and a TypeScript or JavaScript frontend exchange data, before you add or change a request, response, query, event, message, or error…
SkillAFreshness: A★ 1Claude Code +9 - 0.592score
deterministic-tests
futex-ai
Use when a test fails and then passes on a rerun, when a test is flaky or timing-dependent, when a test would assert elapsed time, a timeout, or performance…
SkillAFreshness: A★ 1Claude Code +9 - 0.592score
git-commit
futex-ai
Use before you commit and directly after, when you write a commit message or a pull request title and description, when you squash, when the diff against…
SkillAFreshness: A★ 1Claude Code +9 - 0.592score
rust
futex-ai
Use before you write or change Rust code, add a crate or a dependency to a Cargo workspace, write or move a Rust test, define or map an error type, or add a…
SkillAFreshness: A★ 1Claude Code +9 - 0.592score
implementation-review
futex-ai
Use after you push a branch and the work needs a review, before you run a review or apply its findings, when a plan's review TODO is reached, when the user…
SkillAFreshness: A★ 1Claude Code +9 - 0.592score
writing-skills
futex-ai
Use when you create or edit a skill, write or change a skill description, or when an agent skipped a rule, did not load a skill, or loaded the wrong one.
SkillAFreshness: A★ 1Claude Code +9
Done for you
Rather not set it up yourself?
Chalk Labs scopes, builds and runs agents for your team: the task audit, the eval harness, the integrations and the weekly ops report. Fixed scope, priced before we start.
See how we build AI agents →Read the case studies
Chalk Labs is a two-founder AI and growth studio. We build AI agents, CRMs and MVPs for Web3 and AI companies, and have shipped 50+ product and token launches.
- launches shipped
- 50+
- launches shipped
- media placements
- 500+
- media placements
- founders on every account
- 2
- founders on every account
Press we've placed: Forbes · CoinDesk · Decrypt · Cointelegraph · The Block
Rahil Jain, Co-founder, Chalk Labs. Ex-Goldman Sachs engineer. You talk to the person who builds it.
Get finish-work running in your business
Tell us what you want to automate. A founder replies within one working day with a plan, not a pitch deck.
Cite this page
finish-work skill. Agent Mart. https://agentmart.dev/skills/futex-ai-finish-work (last updated 2026-10-24)