Skip to content
Back to skills

New Task

ASecurity

Orient a new application task and choose a minimal TODO, a small maker/reviewer pair, or a real dependency graph; preserve product checks and existing approvals without contract ceremonies.

  • 4 stars
  • 0 votes
  • 0 copies
  • 0 views
  • Added November 7, 2026
ai-agentsgosecurity

Security analysis

A100/100

Pro scans all 2 files and shows the line behind each finding

Scanned November 7, 2026

npx -y skills add PapeMati98/Ai-Stack --skill new-task --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of New Task?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for New Task
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/papemati98-new-task/badge)](https://www.skillsdirectory.com/skills/papemati98-new-task)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: new-task
description: Orient a new application task and choose a minimal TODO, a small maker/reviewer pair, or a real dependency graph; preserve product checks and existing approvals without contract ceremonies.
---

# New task

1. Read project AGENTS.md and the relevant APP_PROFILE/active manifest if present.
   Use existing verified state; do not reread unrelated project documents.
2. Apply select-skills once for a new goal or a material change of scope, domain,
   risk, QA or inventory. Reuse the selection and relevant current context for
   unchanged follow-ups and status. Select 0-4 needed specialists semantically
   from the complete current inventory; a shortlist is only a hint.
3. The current strong parent plans once. A cheap parent requests one bounded plan
   from a qualified model. Do not duplicate planning or request a planning signature.
4. Simple response or small local change: inline TODO and proportional verification.
   Small delegated change: one todo.json and agent-loop's maker/reviewer pair through
   agent-graph/scripts/todo_pair.py; no graph scheduler. Larger dependent or parallel
   work: one todo.json through todo_graph.py. Each delegated task has scope, criteria,
   executable checks, dependencies and an independent model lineage for review.
5. Include security/data checks in the same task where relevant. For changed UI use
   design-ui. Use design-loop for a new visual decision, changed layout, navigation
   or primary journey, or a redesign. Direct-build reuses a confirmed pattern, even
   for a new screen, with independent UAT. Size alone does not trigger an artboard.
   Define roles, expected/actual and
   an isolated UAT environment before coding; unit/build alone do not accept a UI path.
6. Read Hub context once in the parent, pass only relevant facts to workers. Use
   Office presence and private/task-channel reads at entry and stage checkpoints.
   The task channel supplied by the leader takes priority. Do not add empty polling.
7. Integrate and verify the actual product. For material deliverables keep a concise
   status: function | exists | tested | integrated | missing. Do not confuse preserved
   artifacts or successful transport with product acceptance or a Hub receipt.
8. Preserve useful code when only its report fails. Repair transport deterministically,
   prove model/workspace/base/scope/hashes, run checks and independent review. Never
   mark ACCEPT because files exist or ask a maker to redo unchanged work for formatting.

Existing no-delete, privacy, production consent and backup rules remain binding.
Routine local delegation needs no new approval. Never invent legal decisions.
Every session leaves a canonical Hub closeout; failure is knowledge_closeout: BLOCKED.
Shared skill changes require private backup PUSHED/CURRENT and a narrow rollback.

Historical contract commands remain in references/legacy-recovery.md, HISTORICAL ONLY.
Do not load that document or run init/approve for new TODO work.

Files in this skill

  • SKILL.md3 KB
  • references/legacy-recovery.md4.9 KB

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.