agentleFS
wernerbattwernerbatt/obsidian-gtd-cli/.claude/skills/next/SKILL.md

next

Skill0 starsChanged 7 months agoMITView on GitHub

AI-driven task picker that selects the single best thing to work on, pre-fetches context, assesses delegation level, and helps execute. Triggers: "what's next", "what should I work on", "next", "pick a task", "I'm free", or any request for the next action.

40 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use quick, status and 7 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • codebycoreySkill

    quick

    Quick start a change with minimal interaction. Use when fast-tracking a straightforward change that needs proposal, design, and tasks in one step.

    Also in 3 projects that use this

  • AnthropicSkill

    status

    Case status summary by audience — client-facing (plain language), internal (for the professor), or court-ready (formal caption format per local rules). Same facts, different framing and depth. Use when a student needs to update the client, brief the professor, or prepare a court status report.

    Also in 9 projects that use this

  • MicrosoftSkill

    cleanup

    Also in 4 projects that use this

  • genomewalkerSkill

    explore

    RLM-style recursive memory exploration - dynamically navigate the memory graph

    Also in 3 projects that use this

  • GoogleSkill

    critique

    Expertise in auditing and fixing repository scripts and GitHub Actions workflows to ensure technical robustness and security.

    Also in 3 projects that use this

  • OpenAISkill

    audit

    Audit or critique a product flow, journey, workflow, funnel, onboarding path, checkout path, settings path, screen, or multi-step product experience by capturing screenshots first, then reporting UX, design, and accessibility findings inline from that evidence. Use Figma only when the user explicitly asks for a board. Use when the user asks to audit, review, critique, inspect, assess, analyze, evaluate, or give feedback on a product experience.

    Also in 7 projects that use this

The file

Show its rules and text

We split this skill into the rules it gives an agent

31 rules. 0 are common, 27 are distinctive, and 4 are very specific to this project.

  • 0

    Common

    Most of the 36 edited versions we read keep these.

  • 27

    Distinctive

    This author's own choices. Worth a look.

  • 4

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    AI picks, you veto: No lists. One task at a time.

    line 16
  2. Distinctive:

    Read/search vault files

    line 39
  3. Distinctive:

    Create a draft or note (can be deleted)

    line 40
  4. Distinctive:

    Add a task to the daily note (can be removed)

    line 41
  5. Distinctive:

    Look up calendar events

    line 42
  6. Distinctive:

    Delete or move vault content

    line 48
  7. Distinctive:

    Session momentum: if user just completed 2+ @quick tasks, suggest a @deep task ("you've got momentum — time for the real work?")

    line 85
  8. Distinctive:

    Recency: prefer tasks from recent daily notes and active projects over deep archive

    line 86
  9. Distinctive:

    Before presenting, gather everything the user needs:

    line 90
  10. Distinctive:

    Read the vault note if one exists (wikilink in task)

    line 92
  11. Very specific:

    Read the project file if the task is in a GTD/Projects/ file

    line 93
  12. Distinctive:

    Check for related tasks (same 🆔 chain)

    line 94
  13. Distinctive:

    Check calendar for scheduling constraints

    line 96
  14. Distinctive:

    Keep it tight. No essays.

    line 114
  15. Very specific:

    When a task is done/dropped/skipped, immediately present the next one.

    line 187
  16. Distinctive:

    If the user has completed 3+ tasks in a row, acknowledge the momentum:

    line 189
  17. Distinctive:

    When detected, gently surface it:

    line 201
  18. Very specific:

    If yes → hand off to the /stuck skill.

    line 204
  19. Very specific:

    If the user engages with a task but hits resistance (can't start, hesitating, avoidance language), switch to /stuck mode seamlessly.

    line 213
  20. Distinctive:

    When a meeting is ≤30 min away:

    line 217
  21. Distinctive:

    Don't present lists. Ever.

    line 231
  22. Distinctive:

    Don't ask "what's your energy level?" — infer from time of day and session pattern.

    line 232
  23. Distinctive:

    Don't over-explain the delegation level — just act on it.

    line 233
  24. Distinctive:

    Don't ask permission for reversible actions — do them, report back.

    line 234
  25. Distinctive:

    Don't stall. If context fetch takes too long, present the task with what you have.

    line 235
  26. Distinctive:

    Don't repeat a skipped task in the same session unless the user asks.

    line 236
  27. Distinctive:

    Always include the link when presenting a URL-based task.

    line 237
  28. Distinctive:

    Use gmcli to fetch email content for Gmail-linked tasks — don't say "I can't access it."

    line 238
  29. Distinctive:

    If invoked with no context:

    line 242
  30. Distinctive:

    Check calendar for imminent meetings (next 30 min)

    line 244
  31. Distinctive:

    No questions. No "what would you like to focus on?" Just go.

    line 248

“Common” is measured: kept by at least 6 in 10 of the 36 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

/next — Task Picker & Executor

One command. AI picks the task, assesses how much of you it needs, helps execute, moves on.

Core Philosophy

  • AI picks, you veto. No lists. One task at a time.
  • Every task has a delegation level. The question isn't just "what" but "how much of you does it need?"
  • Reversibility is the safety line. AI can do anything reversible without asking. Always report what was done.
  • Motion over perfection. Picking a good task fast beats picking the perfect task slowly.
  • Stay with the task. The goal is to COMPLETE the task, not whizz through the list. Stay with a task until it's done or the user explicitly says skip/next.

Delegation Levels

Based on the Teresa Torres framework. Assessed per task before presenting.

LevelWhat happensYou doAI does
AutomateAI does it entirely, reports backRead the summaryEverything
AssistAI drafts/researches, you reviewApprove or tweakThe heavy lifting
PromptAI needs one quick input from youAnswer one questionEverything else
CollaborateRequires your judgment, creativity, or emotionThink and decideSupport, scaffold, unblock

Reversibility Guide

These actions are reversible — AI can do them and report after:

  • Mark a task done (can be unchecked)
  • Web research and summarise findings
  • Read/search vault files
  • Create a draft or note (can be deleted)
  • Add a task to the daily note (can be removed)
  • Look up calendar events

These need your go before executing:

  • Send a message or email
  • Book or create calendar events
  • Delete or move vault content
  • Edit existing task descriptions
  • Anything involving another person

Always tell the user what you did, even for reversible actions. One line is enough.

Vault & Calendar Operations

Use the /obsidian skill for all vault reads and writes. → .claude/skills/obsidian/SKILL.md

Use gccli for calendar operations.

Task Selection Algorithm

Step 1: Gather candidates

Pull in parallel:

  1. Calendar — events in the next 2 hours (via gccli). If a meeting is within 30 min, it takes priority as "prep for this meeting."
  2. Vault tasks — all incomplete, non-someday (⏬), non-blocked (⛔) tasks via Dataview eval.

Step 2: Score and rank

Pick the single best task using this priority:

  1. Meeting in ≤30 min → prep for it (always wins)
  2. 🔺 Highest priority → do first
  3. ⏫ + @quick → momentum builders
  4. ⏫ + @deep → the real work
  5. 🔼 overdue or aging >7 days → surface with Avoider nudge (see PQ integration)
  6. 🔼 @quick → easy wins
  7. Normal priority, due today → obligations
  8. @quick with no priority → filler if energy is low

Context signals to factor in:

  • Time of day: morning → favour @deep; afternoon → favour @quick; evening → favour @batch/@read
  • Session momentum: if user just completed 2+ @quick tasks, suggest a @deep task ("you've got momentum — time for the real work?")
  • Recency: prefer tasks from recent daily notes and active projects over deep archive

Step 3: Pre-fetch context

Before presenting, gather everything the user needs:

  • Read the vault note if one exists (wikilink in task)
  • Read the project file if the task is in a GTD/Projects/ file
  • Check for related tasks (same 🆔 chain)
  • Resolve URLs — for links, fetch the page title/summary
  • Check calendar for scheduling constraints

Step 4: Assess delegation level

SignalDelegation level
Task is a URL to read/watchAutomate (summarise) or Assist (summarise + recommend). Check Clippings/ folder first — if a clipping exists, read it. If not, ask the user to create one via the snipping tool, or try gmcli for email content.
Task is "research X"Automate (search + summarise) or Assist (present findings)
Task is "check on X" / recurring checkAutomate (check + report)
Task is "draft X" / "write X"Assist (draft, you review)
Task is "discuss with [person]"Prompt (AI preps talking points, you have the conversation)
Task is "decide X" / @ponderablesCollaborate (explore together)
Task is @stuckCollaborate (trigger stuck skill)
Task involves sending/booking/committingAssist (AI preps, you approve the action)
Task is vague / no clear next actionPrompt (clarify what it actually is, then re-assess)

Presentation Format

Keep it tight. No essays.

[task description]
📍 [file path]  [delegation level icon]

[1-2 lines of pre-fetched context]

[action prompt based on delegation level]

By delegation level:

Automate:

Check on mollycantillon @quick
📍 Daily/2026/2026-01-25.md  🤖 Automate

Checked — last post was 2 weeks ago, nothing new since.

✅ Marked done.

(Already did it. Just informing.)

Assist:

Research Aeyla sleep solutions @deep 🔼
📍 Daily/2026/2026-01-11.md  🔧 Assist

Aeyla sells weighted blankets and cooling pillows. 4.6★ on Trustpilot,
30-night trial, £79-£149 range. Main product is the dual-comfort pillow.

Want me to write up a comparison note, or is this enough to decide?

Prompt:

Discuss putting password in sealed envelope with Sharné @sharne
📍 GTD/Projects/List of accounts.md  💬 Prompt

You have a password vault. The task is about emergency access if
something happens to you.

When are you next seeing Sharné today/tonight? I'll prep the talking points.

Collaborate:

Werner's AI Career Strategy @ponderables 🔺
📍 Daily/2026/2026-02-21.md  🤝 Collaborate

This has been 🔺 for a month. Your vault note explores three paths:
stay technical, move to AI product, or independent consulting.

Where's your head at with this right now?

Flow

User: "what's next"

AI: [presents one task with context + delegation assessment]

User: "go"     → AI executes at the assessed delegation level
      "skip"   → AI picks the next task (no list, no guilt)
      "done"   → AI marks done, picks next
      "drop"   → AI marks dropped, picks next
      [talks]  → Collaborative mode — work through it together

After completion

When a task is done/dropped/skipped, immediately present the next one. No "what would you like to do?" — just keep the conveyor belt moving.

If the user has completed 3+ tasks in a row, acknowledge the momentum:

"That's [N] knocked off. Keep going or take a break?"

PQ Integration

Avoider Detection

Watch for these signals:

  • User says "skip" 3+ times in a row
  • User says "not the right time" or "I'll come back to this"
  • 🔼 tasks that have been aging >7 days keep getting skipped

When detected, gently surface it:

"That's 3 skips. Your Avoider might be running — want to look at what's making these feel heavy?"

If yes → hand off to the /stuck skill.

When presenting 🔼 tasks aging >7 days

Note the pattern without lecturing:

"This has been sitting since [date]. Might be worth a look — or drop it if the moment's passed."

Stuck Handoff

If the user engages with a task but hits resistance (can't start, hesitating, avoidance language), switch to /stuck mode seamlessly. Don't announce "switching to stuck skill" — just start coaching.

Calendar Integration

When a meeting is ≤30 min away:

Meeting: 1:1 with Craig (2:00 PM, 30 min)
📍 Google Calendar  🔧 Assist

Last met: Mar 14 — discussed API spec review.
You have 2 @waiting tasks mentioning Craig.

Want me to pull those tasks and draft an agenda?

What NOT to Do

  • Don't present lists. Ever. One task.
  • Don't ask "what's your energy level?" — infer from time of day and session pattern.
  • Don't over-explain the delegation level — just act on it.
  • Don't ask permission for reversible actions — do them, report back.
  • Don't stall. If context fetch takes too long, present the task with what you have.
  • Don't repeat a skipped task in the same session unless the user asks.
  • Always include the link when presenting a URL-based task.
  • Use gmcli to fetch email content for Gmail-linked tasks — don't say "I can't access it."

Cold Start

If invoked with no context:

  1. Check calendar for imminent meetings (next 30 min)
  2. Pull top-priority tasks from vault
  3. Present the best one

No questions. No "what would you like to focus on?" Just go.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.