haksolot/ank/plan

Claude Code and agent skill by haksolot/ank

Trust tier 2
  • community
LLM review pending
Source check pending

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

Interview a goal into decisions and tasks recorded in .ank/. Use when someone brings a feature, change, or problem to plan before implementation in a repository with a .ank/ directory.

Install

npx vskill@latest install haksolot/ank/plan

SKILL.md

ank-plan

Planning turns an intention into entities the loop can consume: decisions as ADRs, rules as specs, work as tasks, order as blocked_by. The interview is how uncertainty leaves the plan before an executor meets it.

The ank skill is the contract and applies here in full. This file adds the planning policy only.

Read before asking

Never ask a question the repository answers. Before the first question:

ank context <path> what already binds this perimeter ank find <query> what already exists near the goal ank find --type spec the specification itself ank scope <path> everything covering the files in play ank log <id> what previous holders learned

Then read the code the goal touches. A question answered by reading, asked anyway, teaches the human to stop answering.

Re-measure what you are about to plan around. A number from an earlier session is a claim about a tree that has moved since, and a goal built on one is a goal aimed at the wrong place. Take the measurement again before the first question: the work may already be done, and where it is not, the shape of what remains is rarely the shape that was reported.

The interview

Work the frontier: the decisions whose prerequisites are already settled. Everything downstream of an open decision waits.

For each decision on the frontier:

  1. Check the repository has not already made it.
  2. Recommend one option, and say why.
  3. Ask.

One round holds only questions independent of each other. Fold the answers in, recompute the frontier, repeat. Stop when the frontier is empty: nothing important left silently assumed.

A question only running code can answer is not a question. It is a task with an experiment for a criterion.

Write the plan into the corpus

Each settled decision becomes exactly one kind of entity:

a durable architectural decision ank new adr lands proposed a normative rule ank new spec lands proposed concrete bounded work ank new task a scope is mandatory an ordering between tasks ank amend --blocked-by a passing observation ank log

A task names the policy its shape calls for with ank new task --method <name>: diagnose when the criterion names a defect, tdd when it names behaviour to build. Leave it unset otherwise, since an absent method says only that nobody designated one. The name is the sibling's directory, never ank-tdd.

A done_criteria states what a verifier can check, not what an executor should attempt. If the criterion will not write, the decision is not settled: back to the frontier. claim is where it is enforced, refusing at 7; new writes a task with no criterion and says nothing. So a criterion left for later is a task nobody can start, and the plan is short one decision.

Where planning ends

Proposed ADRs and specs bind nobody until a human ratifies them: a signed accept, a ref that binds wherever the file is read once it reaches the remote. Say what is waiting and stop. Then ank graph and ank check on the result: a cycle, an orphan or a finding is a plan not finished.

Security

  • Security scanPassed, 100 of 100Pattern checks on the skill's files.
  • Model reviewNot runA model review is scheduled for the next rescan.

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/haksolot/ank/plan/badge)](https://verified-skill.com/skills/haksolot/ank/plan)

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