linear-task-planning
Skill7 starsChanged 4 months agoMITView on GitHub
Use when generating, decomposing, creating, or updating implementation tasks in Linear through the official Linear MCP. Requires a reviewed preview before every mutation.
In this repository5 other files it gives its agents.
The file
Show its rules and text
We split this skill into the rules it gives an agent
17 rules. About 0 are common, 17 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈17
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 12
Read only the project material needed for the requested scope.
- Distinctive: line 19
Do not generate implementation work that depends on an unresolved decision.
- Distinctive: line 23
Use Linear MCP read tools before proposing work:
- Distinctive: line 27
Read available statuses, labels, milestones, cycles, and issue relationships.
- Distinctive: line 28
Search for matching issue titles and source identifiers to prevent duplicates.
- Distinctive: line 29
Ask the user when team or project placement is ambiguous.
- Distinctive: line 31
Never invent Linear IDs, status names, labels, projects, or relationships.
- Distinctive: line 40
Prefer a vertical, observable outcome over vague layer activity.
- Distinctive: line 42
Do not add speculative features, abstractions, or compatibility work.
- Distinctive: line 61
Before creating or updating anything, show:
- Distinctive: line 70
Ask for explicit approval.
- Distinctive: line 78
Create issues with the approved descriptions and placement.
- Distinctive: line 79
Add relationships only when supported by the source plan.
- Distinctive: line 80
Do not assign owners unless explicitly requested.
- Distinctive: line 81
Do not silently alter existing issue scope or status.
- Distinctive: line 83
If a mutation partially fails, stop, report what succeeded with URLs, and propose the smallest recovery action.
- Distinctive: line 87
Read the resulting Linear objects and confirm:
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Linear Task Planning
Turn approved engineering plans into clear, self-contained Linear issues. Use the official Linear MCP for discovery and mutations. Do not implement code or edit project files.
Source Authority
Read only the project material needed for the requested scope. Resolve conflicts in this order unless the project says otherwise:
- Accepted architecture decisions.
- Architecture and component designs.
- Roadmap phase gates and phase plans.
- The user's current instruction.
Do not generate implementation work that depends on an unresolved decision. Create a decision or research issue only when the source plan explicitly allows it or the user approves it.
Discover Linear Context
Use Linear MCP read tools before proposing work:
- Identify the workspace and candidate team.
- Find an existing project before proposing a new one.
- Read available statuses, labels, milestones, cycles, and issue relationships.
- Search for matching issue titles and source identifiers to prevent duplicates.
- Ask the user when team or project placement is ambiguous.
Never invent Linear IDs, status names, labels, projects, or relationships.
Decompose Work
Generate tasks from one approved phase, work package, or implementation plan at a time.
- Target roughly half a day to three engineering days per issue.
- Split work above five days or work containing independent contracts.
- Preserve prerequisite order and capability gates.
- Prefer a vertical, observable outcome over vague layer activity.
- Separate reusable harnesses, research decisions, persisted-format changes, and migrations when independently deliverable.
- Do not add speculative features, abstractions, or compatibility work.
Each issue must be implementable without hidden decisions and contain:
- Outcome.
- Scope.
- Non-goals.
- Prerequisites and dependencies.
- Affected contracts or components.
- Proposed implementation steps when established by source material.
- Failure cases.
- Required tests and benchmarks.
- Documentation impact.
- Acceptance evidence.
Include project-specific correctness requirements from contributor guidance and design docs. For database work, this may include crash recovery, deterministic concurrency, branch reachability, denied security paths, compatibility fixtures, or correctness-oracle benchmarks.
Preview Before Mutation
Before creating or updating anything, show:
- Target team and project.
- Proposed issue titles and concise outcomes.
- Initial statuses, labels, and milestone or cycle placement.
- Parent, sub-issue, blocking, and related relationships.
- Existing issues that will be reused or updated.
- Ambiguities or decisions requiring review.
Ask for explicit approval. Read-only discovery does not require approval; every create or update does.
Write Idempotently
After approval:
- Recheck matching issues immediately before creation.
- Reuse or update only confident matches.
- Create issues with the approved descriptions and placement.
- Add relationships only when supported by the source plan.
- Do not assign owners unless explicitly requested.
- Do not silently alter existing issue scope or status.
If a mutation partially fails, stop, report what succeeded with URLs, and propose the smallest recovery action. Do not retry creates blindly.
Verify
Read the resulting Linear objects and confirm:
- Expected issue count.
- Titles and descriptions.
- Team, project, status, labels, and milestone or cycle.
- Parent and dependency relationships.
- No obvious duplicates.
Return a concise summary with issue URLs, skipped work, unresolved questions, and any verification gaps.
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
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.

