Skip to content
Back to search

eliptikwu

MarketplaceWorks with Claude CodeGitHub

Claude Code skills by EliptikWu (LK Industries).

Safety scan

A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.

Checked today
Badge for your README
findsafeskills safety rating

Markdown

[![findsafeskills safety rating](/api/listings/eliptikwu--next-task/badge.svg)](/listings/eliptikwu--next-task)

HTML

<a href="/listings/eliptikwu--next-task"><img src="/api/listings/eliptikwu--next-task/badge.svg" alt="findsafeskills safety rating" /></a>

Install

Install instructions from the README

As written by the repo's author. findsafeskills has not verified these steps.

**Claude Code, as a plugin:**

/plugin marketplace add EliptikWu/next-task
/plugin install next-task@eliptikwu

Or copy `skills/next-task/` to `~/.claude/skills/next-task/` (every project) or to `<repo>/.claude/skills/next-task/` (one project).

**Other tools:** copy `skills/next-task/` to `<repo>/.agents/skills/next-task/` for one project, or to `~/.agents/skills/next-task/` for all of them. Every tool below reads that folder.

| Tool | Run it | Fresh session | |---|---|---| | Codex | `$next-task` | `/new` | | Gemini CLI | say "next task" (it picks the skill itself) | `/clear` | | GitHub Copilot in VS Code | `/next-task` | New Chat (Ctrl+N) | | GitHub Copilot CLI | `/next-task` | `/clear` | | Cursor | `/next-task` | New chat (Cmd/Ctrl+N) | | Devin Desktop (formerly Windsurf) | `/next-task` | a new conversation |

The project's instructions go in whatever file your tool reads (`CLAUDE.md`, `AGENTS.md`, `GEMINI.md`...). The skill follows that file.

If a project already has its own skill for this loop, that one wins.

No backlog yet? The skill offers to create one. The format is in [`skills/next-task/reference/backlog-format.md`](skills/next-task/reference/backlog-format.md).

What was scanned

The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.

.claude-plugin/marketplace.json · Manifest · 16 lines
  1. {
  2. "name": "eliptikwu",
  3. "description": "Claude Code skills by EliptikWu (LK Industries).",
  4. "owner": {
  5. "name": "EliptikWu"
  6. },
  7. "plugins": [
  8. {
  9. "name": "next-task",
  10. "source": "./",
  11. "description": "One task per session, then /clear: measured 60 % fewer tokens per PR without skipping tests or review.",
  12. "version": "1.0.0",
  13. "license": "Apache-2.0"
  14. }
  15. ]
  16. }
README.md · README · 187 lines
  1. <img src=".claude-plugin/icon.png" width="96" alt="">
  2. # next-task
  3. [Leer en español](README.es.md)
  4. An agent skill for working through a backlog **one task per session**. It uses the open
  5. [Agent Skills](https://agentskills.io) format, so it runs in Claude Code, Codex, Gemini
  6. CLI, GitHub Copilot, Cursor and Devin Desktop (formerly Windsurf). Each session it
  7. picks the next item and takes it to a verified pull request. It leaves a one-line handoff
  8. in the repo, then tells you to start a fresh session (`/clear`) and type `/next-task` again.
  9. You type two commands. Each task costs 60 % fewer tokens, without skipping tests or review.
  10. ## Measured savings
  11. These numbers come from 561 real pull requests in two production projects:
  12. - [LK Creator](https://lkcreator.com), a FastAPI + Next.js SaaS that converts custom
  13. Minecraft content.
  14. - ISO 21101 turismo, a safety-management web app for adventure-tourism operators.
  15. The data is the token usage that Claude Code already stores in its local session
  16. transcripts. Same developer, same model (Claude Opus), November 2026. Other tools were not
  17. measured: the idea should carry over, since they all re-read the conversation on every
  18. step, but we have no numbers for them.
  19. | | One task per session | Chained tasks (≥3 PRs in one session) | Savings (95 % CI) |
  20. |---|---|---|---|
  21. | Cost per PR, LK Creator | $4.39 (162 PRs) | $14.58 (131 PRs) | **70 %** (37–72 %) |
  22. | Cost per PR, ISO 21101 turismo | $5.10 (49 PRs) | $9.63 (219 PRs) | **47 %** (18–78 %) |
  23. | **Cost per PR, both** | | | **60 %** (42–74 %) |
  24. | **Context per model call, both** | 123–130k tokens | 342–357k tokens | **73 %** (62–75 %) |
  25. **On a subscription.** We read `/usage` three times, around two work periods, and
  26. summed the tokens of every session in between, across all projects and including
  27. subagents.
  28. - One point of the weekly limit cost $17.4 of API-equivalent tokens: 11 points in total,
  29. 0.059 and 0.057 points per dollar in the two periods.
  30. - `/usage` rounds to whole points, so this figure is ±9 %.
  31. - So one task per session uses about **0.25–0.29 % of the weekly limit per PR**. Chaining
  32. tasks uses **0.55–0.84 %**. That is roughly **twice as many PRs per week** on the same
  33. plan.
  34. **Why it works.** Every model call re-reads the whole conversation. In a chained session
  35. the context per call keeps growing; it passed 500k tokens after about 10 tasks. So each
  36. later task pays for all the earlier ones. A fresh session starts every task with the same
  37. small context.
  38. ### Quality: no worse, and the safety steps actually happen
  39. We joined the first 510 of these PRs with their GitHub history:
  40. | | One task per session | Chained | Difference (95 % CI) |
  41. |---|---|---|---|
  42. | Merged | 98.2 % | 98.6 % | none (−3.5 to +3.1 pts) |
  43. | Reverted later | 0 % | 0 % | none |
  44. | Tests / typecheck run before opening the PR | **97.6 %** | 75.1 % | 22.5 pts more (2 to 38) |
  45. | Code review run before opening the PR | **87.9 %** | 14.8 % | 73 pts more (55 to 87) |
  46. In long chained sessions the model quietly starts skipping verification and review as the
  47. context fills up. A fresh session follows the full checklist almost every time.
  48. We also measured "rework within 7 days": a later `fix` PR that touches the same code files.
  49. It favored one task per session on ISO 21101 turismo (8.6 % vs 59.2 %), but it is too
  50. noisy to claim. In these repos a third of all PRs are planned `fix` work on the same
  51. modules, so the combined interval crosses zero.
  52. ### Behavior tests
  53. Before release we ran the skill and its previous version on three scripted repos, graded
  54. automatically (scripts in [`tests/`](tests/)):
  55. - **a real bug first:** fix it, one task only;
  56. - **an already-fixed item first:** check the premise, mark it obsolete, move on;
  57. - **no backlog:** offer to create one and don't invent work.
  58. This version passed 19/19 checks with the same token use as the previous one (≈54k per
  59. run). The previous version wrote invented PR numbers into the handoff when there was no
  60. remote, and this version fixes that.
  61. ### How it was measured
  62. The scripts are in [`bench/`](bench/).
  63. - **Cost.** We took the tokens of every model call, including subagents, and weighted them
  64. by Anthropic's price ratios: cache read 0.1×, cache write 1.25× or 2×, output 5×. We
  65. checked these weights against the cost Claude Code records, and they agree within ±10 %.
  66. - **Task.** A task is a pull request actually created in that session: a `gh pr create`
  67. that returned a PR URL.
  68. - **Confidence intervals.** We ran a 5,000-sample bootstrap over **sessions**, not tasks.
  69. That way one long session doesn't count as 100 independent samples.
  70. ### Limits, stated plainly
  71. - It's observational, not an A/B test. The tasks in the two groups are different, and the
  72. chained group has only 11 sessions.
  73. - The dollar figures are API-price equivalents. The subscription figure comes from two
  74. periods and should be read as an estimate.
  75. - The quality table shows that checks were *run*. It cannot show that each one was run
  76. well.
  77. ### Run it on your own Claude Code sessions
  78. ```bash
  79. python bench/measure.py ~/.claude/projects/<your-project-folder> next-task sessions.json
  80. python bench/compare.py sessions.json
  81. python bench/bootstrap.py "mine=sessions.json"
  82. ```
  83. ## What it does in each session
  84. 0. **Guard.**
  85. - If this conversation already closed a task, it asks you to `/clear` first.
  86. - If one of your PRs is still open, it finishes that one first.
  87. - If the default branch's CI is red, fixing it *is* the task.
  88. 1. **Pick.** It takes the first open item of the backlog (`BACKLOG.md`, `TODO.md`, or
  89. GitHub issues), reading only the part it needs.
  90. 2. **Check the premise.** It confirms in the code that the problem still exists, and drops
  91. items that are already fixed.
  92. 3. **Do it.** It finds the root cause and fixes it in the shared mechanism, in the
  93. surrounding style.
  94. 4. **Verify.** It runs the checks CI would run, then reviews the diff.
  95. 5. **Open the PR.** It merges only if the project's instructions file explicitly allows it, and only when
  96. checks pass.
  97. 6. **Hand off.** It writes one line in the backlog, with the PR number and the one thing
  98. the next session must know.
  99. 7. **Stop.** It tells you to start a fresh session (`/clear`), then `/next-task`.
  100. It follows these token rules throughout:
  101. - the repo is the memory;
  102. - it reads line ranges, not whole files;
  103. - wide searches go to a subagent that answers in under 15 lines;
  104. - every command's output is trimmed;
  105. - it uses CLI tools over MCP;
  106. - it never re-derives a fact it already established;
  107. - it stops after two failed attempts at the same fix;
  108. - compaction is treated as a sign that the task was too big.
  109. It works without a GitHub remote. In that case it commits on a branch and leaves the PR to
  110. you.
  111. It answers in your language; the skill itself is written in English because that's what
  112. the model reads.
  113. ## Install
  114. **Claude Code, as a plugin:**
  115. ```
  116. /plugin marketplace add EliptikWu/next-task
  117. /plugin install next-task@eliptikwu
  118. ```
  119. Or copy `skills/next-task/` to `~/.claude/skills/next-task/` (every project) or to
  120. `<repo>/.claude/skills/next-task/` (one project).
  121. **Other tools:** copy `skills/next-task/` to `<repo>/.agents/skills/next-task/` for one
  122. project, or to `~/.agents/skills/next-task/` for all of them. Every tool below reads that
  123. folder.
  124. | Tool | Run it | Fresh session |
  125. |---|---|---|
  126. | Codex | `$next-task` | `/new` |
  127. | Gemini CLI | say "next task" (it picks the skill itself) | `/clear` |
  128. | GitHub Copilot in VS Code | `/next-task` | New Chat (Ctrl+N) |
  129. | GitHub Copilot CLI | `/next-task` | `/clear` |
  130. | Cursor | `/next-task` | New chat (Cmd/Ctrl+N) |
  131. | Devin Desktop (formerly Windsurf) | `/next-task` | a new conversation |
  132. The project's instructions go in whatever file your tool reads (`CLAUDE.md`, `AGENTS.md`,
  133. `GEMINI.md`...). The skill follows that file.
  134. If a project already has its own skill for this loop, that one wins.
  135. No backlog yet? The skill offers to create one. The format is in
  136. [`skills/next-task/reference/backlog-format.md`](skills/next-task/reference/backlog-format.md).
  137. ## License
  138. [Apache 2.0](LICENSE). You can use, change and share it. Copies and derived works must
  139. keep the copyright and the [NOTICE](NOTICE) file, and must say what was changed. The
  140. license doesn't grant use of the author's name to promote a derived work.
  141. Made by [EliptikWu](https://github.com/EliptikWu) at LK Industries.

Ratings and comments

Sign in to rate this skill.

Sign in

Free to use. If it’s helped, pass it on.