raoofaltaher/ultrapowers/task

Trust tier 3
  • community
Source check pending

Checking that the SKILL.md file is still in the repository.

Use when your human partner names a ticket id and wants to know where it stands, resume it, or load its documents, before doing any work on it

Install

npx vskill@latest install raoofaltaher/ultrapowers/task

SKILL.md

Task

Overview

Load every markdown document a ticket has and report where it stands. This is a bulk loader, not a summary: cost is not a reason to read less. It writes nothing and starts nothing.

Announce at start: "I'm using the task skill to load ticket <ID>."

Arguments

ticket (required): a ticket id, or the task URL of an Odoo source. Your argument, as the harness passed it: $ARGUMENTS. A single word there is the ticket. A task URL is resolved first: node "<NEW_TASK_DIR>/scripts/fetch-ticket.mjs" resolve "<URL>" (the new-task skill directory beside this one) prints the id to use for every step below. Only when those backticks are empty, or still hold the unreplaced placeholder (a dollar sign followed by the word ARGUMENTS), did the harness not pass it: then read it from the trailing ARGUMENTS: line of the message that invoked this skill.

With no ticket, stop and print usage: /ultrapowers:task <ticket>.

Step 1: Manifest

<SKILL_DIR> is the directory this SKILL.md was loaded from; <ID> the ticket. Fill both in before running.

bash
bash "<SKILL_DIR>/scripts/manifest.sh" "<ID>"

Sections: MARKDOWN TO READ (path, bytes, estimated tokens), STAGE LOG (the autopilot engine's stage, status, mode, chain check and last ten log lines, or (none) for a ticket that never ran under it), NON-MARKDOWN (listed, not read), REPO STATE (per repo: branch, whether it is a ticket branch, last ten commits, short status). Read the output before continuing. Any line containing ERROR means stop and print it verbatim; if the project is not scaffolded, offer /ultrapowers:init.

Step 2: Read everything listed

With the file-reading tool, one call per file, in this order: tasks/, specs/, plans/, reviews/. That runs brief, design, plan, outcome. Not through the shell: large output is diverted and only a preview reaches you. Read all of them; the point is that your human partner never names a file.

Step 3: Report

In this order:

  1. Position, one sentence. What stage is the ticket at? When the STAGE LOG section is not (none), its stage and status are the position; say so, name the packet or pull request it lists, and note a broken chain as a finding. Otherwise infer the position from the documents.
  2. Done, with evidence from the documents.
  3. In flight: uncommitted files, an open review finding, a ticket branch with work on it.
  4. Next step, in order, on the ticket route: brief, spec, plan, the branch with its work, reviews/<ID>/TASK-REVIEW.md (the whole-ticket review with the vulnerability hunt and the QA gate), then the finish. A ticket whose plan tasks are all done and whose branch has no TASK-REVIEW.md is at the whole-ticket review.
  5. Every missing part, by name. ABSENT folders are normal early on; say which. When all four are absent, say the ticket does not exist yet and suggest /ultrapowers:new-task <ID>.
  6. The non-markdown list, when it is not empty: the files exist, you did not read them, name one by path to have it read.

Step 4: Write nothing

Do not edit, create, commit or push. Do not start the work the documents describe. If your human partner wants work done, they will say so in their next message.

Checklist

  1. Read the argument; stop with usage when it is missing
  2. Run the manifest; stop on ERROR
  3. Read every listed markdown file, in order
  4. Report the six parts
  5. Change nothing

Red Flags

ThoughtReality
"The plan is 40 KB, I'll skim the headings"Read all of it. The loader exists so nothing is skipped.
"I'll fix that typo in the spec while I'm here"Report it under in flight or next step. Writing is a separate request from your human partner.
"No spec yet, so I'll start brainstorming"Name the gap and suggest the skill. Starting work is your human partner's call.
"I'll cat the files in one shell call"Truncated to a preview. One file-reading call per file.
"It is a one-word, uncommitted change, not really a write"Any change to a file is a write. The loader writes nothing; list the fix under next steps and offer it as a separate action.

Security

  • Security scanPassed, 95 of 100Pattern checks on the skill's files.
  • Model reviewPassed

Quality evaluation

  • Task evaluationNot runNo evaluation yet. Run one in Skill Studio.

Source

Popularity

7-day trend score
5

Not refreshed yet.

Works with

Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.

  • Claude Code
  • Cursor
  • GitHub Copilot
  • Windsurf
  • Codex
  • Gemini CLI
  • Cline
  • Amp
  • Roo Code
  • Goose
  • Continue
  • Kilo Code
  • Junie
  • Trae
  • OpenHands
  • Qwen Code
  • AI Agent
  • Tabnine Chat
  • Sourcegraph Cody
  • Amazon Q
  • Firebase Genkit
  • Sweep
  • Devin
  • Phind
  • Mentat
  • Claude Desktop
  • Aider
  • GPT Pilot
  • Plandex
  • bolt.new
  • v0
  • Replit Agent
  • Anse
  • TypingMind
  • BetterChatGPT
  • LibreChat
  • Warp AI
  • Pieces
  • Fig AI
Embed badge
vskill VERIFIED badge
[![vskill VERIFIED](https://verified-skill.com/api/v1/skills/raoofaltaher/ultrapowers/task/badge)](https://verified-skill.com/skills/raoofaltaher/ultrapowers/task)

Something wrong with this skill or its page? Tell us.