yishan-io/yishan-mono/context-task

Trust tier 3
  • community
Source check pending

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

Use when work needs a tracked Local Task during research, planning, execution, review, or user-directed completion.

Install

npx vskill@latest install yishan-io/yishan-mono/context-task

SKILL.md

Context Task

Use this skill to keep one Local Task accurate. The SQLite-backed Local Task daemon owns task metadata and status. Pi uses the managed daemon through its configured endpoint.

New tasks receive daemon-generated UUIDs. Imported IDs stay opaque. Do not create, parse, or replace an ID.

Project scope uses YISHAN_PROJECT_ID. If it is set, use only that project. Without it, use global tasks only.

Task Metadata

Use task_read without a document to read the synthetic, read-only brief. The brief contains daemon metadata. Use task_update to change the title, description, new, progressing, or cancelled status, priority, or tags. It cannot set done.

Use task_list to filter by status, priority, workspace, or tags. Use task_search to search with those filters. task_finish is the only tool that can set done.

A tracked Local Task may contain multiple independently deliverable plan subtasks. A plan subtask is an execution unit within the tracked Local Task; its independent deliverability does not by itself require a new Local Task.

When a newly discovered issue is related to the active plan subtask, a non-scope-changing clarification may stay in scope. If the discovery changes that plan subtask's frozen scope, requires a new decision or dependency, or changes assumptions for a later plan subtask, record it in notes and stop the current attempt. The controller decides whether to continue and, if so, must replace the affected scope-changing plan content with a newly frozen brief before resuming. Ask the user when the change affects unapproved behavior, commitments, or tradeoffs. For an unrelated issue, ask the user whether to create a separate Local Task.

Context Documents

The daemon provides the paths for three context documents:

  • plan contains the current execution plan.
  • notes contains task-specific research and decisions.
  • outcome contains a factual completion summary.

Before execution, every plan subtask in a tracked Local Task requires a controller-created, frozen brief in plan.md. Each brief must state its scope, non-goals, acceptance criteria or check, and a finite execution budget. The controller is the sole authority to create or replace scope-changing plan content. Do not execute a plan subtask without its brief.

Use task_write to replace notes or outcome. Executors may use it in plan only to record execution details that remain within the applicable frozen brief; only the controller may replace scope-changing plan content. Use task_append_note to add research or progress to notes. Keep the plan coherent. Keep notes specific to this Local Task.

Start a new notes document with this compact, editable Markdown structure:

markdown
# Notes
## Summary
## Decisions
## Evidence and Progress
-YYYY-MM-DD — <fact, result, or change>
## Open Questions

Append dated, factual entries under Evidence and Progress. Move resolved material into Summary or Decisions when it becomes important. Do not preserve an old format for its own sake: when the user asks to change the notes, replace or reorganize the Markdown with task_write. Users can also edit notes.md directly in the file editor.

Completion

Use task_finish only when the user explicitly asks to complete the task. After explicit direction that all work is complete, perform the self-checks, prepare a factual outcome, and finish without asking again. The tool writes the outcome and marks the task done.

Do not complete a task because implementation appears complete.

Boundaries

Do not bypass the daemon. Do not create task identifiers. Do not write the synthetic brief. Do not use task_update for completion.

Security

  • Security scanPassed, 90 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
0

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/yishan-io/yishan-mono/context-task/badge)](https://verified-skill.com/skills/yishan-io/yishan-mono/context-task)

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