get-it-done
Pursue an agreed goal to a working outcome without waiting: decompose it, re-prioritise as evidence arrives, branch, gate, get an independent review (a…
Skill by jpantsjoha
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/jpantsjoha-get-it-done
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: agent-building
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 and code. 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/get-it-done
What's inside
- Skill name
get-it-done- Description
- Pursue an agreed goal to a working outcome without waiting: decompose it, re-prioritise as evidence arrives, branch, gate, get an independent review (a different model or a fresh reviewer subagent — the writer never approves), merge, prune, and bank the operator's decisions in STATUS.md instead of blocking on them. Trigger on "get it done", "don't wait for me", "proceed autonomously", or when an epic or roadmap objective is handed over. Not a close-out: this work set is expected to grow.
- License
- Apache-2.0
- Instructions
- 127 lines in SKILL.md · view on GitHub
Files (1)
- SKILL.md
SKILL.md (first lines)
# Get it done **The planning, the talking, the debating are over. The plan is agreed — execute it.** Do not re-open scope, do not re-surface options already ruled on, do not ask for reassurance. The only reason to interrupt is a decision from the short "Theirs, always" list below; everything else is action. The operator has left. **Finish the phase.** Do not stall on anything you can decide, evidence, or work around — and do not silently decide anything that is genuinely theirs. The whole skill is one judgement: **what is mine to call, and what is theirs?** Get that wrong in the timid direction and the work sits idle for hours. Get it wrong in the reckless direction and you spend their money, their credibility, or their consent on a call they never made. ## The boundary — this pursues a goal, it does not close a queue The unit here is a **goal**: an epic, a roadmap objective, a phase, a thing that must end up *working*. The work set is derived from that goal and **is expected to grow** — you decompose, you discover, you re-prioritise as evidence arrives, and you update the delivery tracker as you go rather than at the end. Done means the goal is functional, not that a list is empty. A close-out is the opposite shape: its work set is **frozen** at invocation and never grows, and it ends in a release advisory. If the operator asked you to close out what is in flight, do that job, not this one. Running this skill against a closing set is how a close-out quietly turns into a new sprint. ## The contract 1. **Do not wait.** No "shall I proceed?", no "let me know". If you can act, act. 2. **Bank, don't block.** A decision that is genuinely theirs goes into `STATUS.md` with a recommendation — not into a message that stops the work. 3. **Land it.** A phase is not done because the code is written. It is done when it is merged to `main`, the branch and worktree are pruned, and the tracked documents match reality. 4. **Never fabricate authority.** Blanket "go" covers ordinary work. It does not convert into permission for the four things below. ## Theirs, always — bank these, never assume them Even under an explicit "don't wait for me": - **Money.** Raising a spending limit, a plan, a paid tier, a metered vendor. - **Consent, credentials, retention, permissions.** Anything a user agreed to, anything that reads or keeps their data, anything holding a secret. - **One-way doors.** Tags, releases, published artefacts, external submissions, force-pushes, deleting a ref whose commits are not reachable from the trunk. - **Self-approval of a governance record.** An ADR is never accepted by the agent that wrote it. Draft it, evidence it, and flag it. Everything else — ordinary code, tests, docs, refactors, issue filing, branch hygiene — is yours. Do it. ## When a dependency is missing, route around it and say so A blocked gate is not a reason to stop; it is a reason to substitute and record the substitution. | Unavailable | Substitute | Record | |---|---|---| | Hosted CI (quota, outage, no runner) | Run the local equivalent of **each** required check | Name each check and its local equivalent, with exit codes | | Hosted code review (quota) | An independent reviewer: a different model or a fresh reviewer subagent | Which reviewer, what verdict, what it found | | The first reviewer unavailable | An alternative only within the project's authorised review policy; otherwise stop the affected gate | Availability, authority and unresolved evidence | | A required check that *cannot execute* | Merge on the local evidence if the repo permits it | **Say "past checks that could not run", never "past checks"** | The distinction in that last row is the whole point. A check that failed and a check that could not start are different facts, and conflating them is how an unreviewed merge gets laundered into a reviewed one. **Never route around a guard that exists to stop you.** A denied push, a protected branch, a consent gate — those are the operator's intent expressed in config. Routing around them is permission laundering, not resourcefulness. ## The loop For each unit of work: 1. **Branch.** Never commit to the trunk. 2. **Build**, with the project's own gates run as you go — not saved for the end. 3. **Prove it bites.** Break the source deliberately and confirm the test fails. A green suite you have not seen go red is a hypothesis. 4. **Review independently.** A fresh reviewer context running `pr-reviewer`, with optional provider diversity under local policy; `adversarial-gate` for design. Follow the [Council contract](../pr-reviewer/references/independent-council.md) for bounded loops. **The writer never approves the work.** If the candidate changes after review, the review is void — re-run it. 5. **Merge** using the project's authorised merge strategy after exact-candidate gates. 6. **Prune** only this lane's clean worktree and merged local/remote branch after verifying merged PR state and integration on main. Squash/rebase changes commit IDs; ancestry alone is insufficient. If integration or ownership is uncertain, preserve the work and explain why. Fetch and fast-forward clean main safely; never reset another lane's checkout or discard uncommitted/unintegrated changes. 7. **Reconcile.** STATUS, CHANGELOG, ROADMAP, ADRs, issues. Findings that live only in a chat message evaporate; file them. ## Parallelism Use worktrees and subagents where the work is genuinely disjoint. **Two agents editing the same file is not parallelism, it is a merge conflict with extra steps** — and most "parallel" refactors of one subsystem collide. Say so rather than parallelising for appearances. ## The handover `STATUS.md` is what they read when they return, so write it for someone who was not in the room. It carries: - **What landed**, with evidence — a SHA, an exit code, an issue number. - **The decisions waiting on them**, each with: a recommendation (a verb and an object, never "consider the options"), what it unblocks, **what it costs if they call it wrong**, and whether it is reversible — `cheap`, `costly`, or `one-way`. Sort so the compounding ones are first. - **What you did not do, and why.** Especially anything you declined as theirs. - **What is still open**, including defects you found and chose not to fix. ## Reporting back
From jpantsjoha/ai-native-developer-experience, Apache-2.0.
More from jpantsjoha/ai-native-developer-experience
- 0.641score
using-the-harness
jpantsjoha
Session-start orientation for this team harness.
SkillAFreshness: A★ 13Claude Code +9 - 0.641score
cloud-expert
jpantsjoha
Cloud guardrails for any vendor workload — Google Cloud (GCP, Vertex AI, GKE), AWS (IAM, EKS, Bedrock), Azure (Entra ID, Policy, AKS), Alibaba Cloud (RAM…
SkillAFreshness: A★ 13Claude Code +9 - 0.641score
operating-model-bootstrap
jpantsjoha
Bootstrap or harden a project's operating model for mixed human and AI delivery teams.
SkillAFreshness: A★ 13Claude Code +9 - 0.641score
pr-reviewer
jpantsjoha
Review gate for correctness, reuse, and simplification.
SkillAFreshness: A★ 13Claude Code +9 - 0.641score
cloud-checkup
jpantsjoha
Read-only SRE checkup of any GCP project: deterministic probes of the edge, Cloud Run services, 7-day error logs, Cloud Scheduler, alert policies and uptime…
SkillAFreshness: A★ 13Claude Code +9 - 0.641score
system-performance-guardrail
jpantsjoha
Measure, root-cause, fix, verify and record performance contention on a developer machine: cloud-sync daemons (iCloud Drive, OneDrive, Dropbox, Google Drive)…
SkillAFreshness: A★ 13Claude Code +9 - 0.638score
cost-guardrail
jpantsjoha
LLM and cloud cost awareness — model tiering, token budgets, right-sizing, and when a cheaper model suffices.
SkillAFreshness: A★ 13Claude Code +9 - 0.636score
frontend-performance-audit
jpantsjoha
Measure real frontend performance — Core Web Vitals (LCP, CLS, INP), page-load traces, CPU and network throttling, Lighthouse accessibility/SEO scores, and…
SkillAFreshness: A★ 13Claude Code +9 - 0.636score
github-manager
jpantsjoha
Use when configuring or optimising GitHub repositories for cost-effective, consistent operations — CI triggers, Actions billing, issue tracking, labelling…
SkillAFreshness: A★ 13Claude Code +9 - 0.636score
mcp-server-scaffold
jpantsjoha
Scaffold and govern an MCP server as a bounded tool seam.
SkillAFreshness: A★ 13Claude Code +9 - 0.636score
spec-first-delivery
jpantsjoha
Force plan, HLD, ADRs, and constraints to exist BEFORE any code is generated.
SkillAFreshness: A★ 13Claude Code +9 - 0.636score
sitrep
jpantsjoha
Synthesise a status, standup, or situation-report from current work state.
SkillAFreshness: A★ 13Claude Code +9
Signals
- Stars
- 13
- Forks
- 4
- Open issues
- 2
- Last commit (this folder)
- Oct 16, 2026 (11 days ago)
- Last release
- Oct 16, 2026 (11 days ago)
From GitHub (fetched Oct 24, 2026).
Details
- License
- Apache-2.0
- Website
- jpantsjoha.com
- First seen
- Oct 24, 2026
- Last checked
- Oct 24, 2026
- Source
- GitHubgithub-graphql
More from jpantsjoha
- 0.641score
system-performance-guardrail
jpantsjoha
Measure, root-cause, fix, verify and record performance contention on a developer machine: cloud-sync daemons (iCloud Drive, OneDrive, Dropbox, Google Drive)…
SkillAFreshness: A★ 13Claude Code +9 - 0.641score
pr-reviewer
jpantsjoha
Review gate for correctness, reuse, and simplification.
SkillAFreshness: A★ 13Claude Code +9 - 0.641score
cloud-expert
jpantsjoha
Cloud guardrails for any vendor workload — Google Cloud (GCP, Vertex AI, GKE), AWS (IAM, EKS, Bedrock), Azure (Entra ID, Policy, AKS), Alibaba Cloud (RAM…
SkillAFreshness: A★ 13Claude Code +9 - 0.641score
operating-model-bootstrap
jpantsjoha
Bootstrap or harden a project's operating model for mixed human and AI delivery teams.
SkillAFreshness: A★ 13Claude Code +9 - 0.641score
cloud-checkup
jpantsjoha
Read-only SRE checkup of any GCP project: deterministic probes of the edge, Cloud Run services, 7-day error logs, Cloud Scheduler, alert policies and uptime…
SkillAFreshness: A★ 13Claude Code +9 - 0.641score
using-the-harness
jpantsjoha
Session-start orientation for this team harness.
SkillAFreshness: A★ 13Claude Code +9
Cite this page
get-it-done skill. Agent Mart. https://agentmart.dev/skills/jpantsjoha-get-it-done (last updated 2026-10-24)