writing-plans

What it does

The skill instructs the agent to produce a comprehensive implementation plan from an approved spec or requirements for a new workstream or escalation. It directs the agent to generate an implementation plan that includes scope checks, file maps, bite-sized tasks with explicit file paths, exact commands, and expected outputs, followed by self-review and a save step that assigns an execution route. It also covers when to use a Planless Slice Lane for micro-slices and when to write a full durable plan for new workstreams, with constraints around authority, verification, and risk assessment.

How it works

• If there is an existing parent plan/spec and the task is a micro-slice, emit a Slice Card and update the parent state; do not save a new plan.
• If there is an approved spec/requirements for a new workstream or escalation trigger, write an implementation plan assuming zero context:
  1. Conduct a Scope Check covering fact/assumption/unknown, baseline, Ripple Signal Triage, compatibility boundary, and dual-track needs.
  2. Create a File map listing what files are created/modified with clear boundaries and existing patterns.
  3. Break down tasks into bite-sized steps (2-5 min each) with exact file paths, complete code, exact commands, and expected output.
  4. Include a Self-review covering spec coverage, placeholders, type consistency, compatibility, verification, and dual-track.
  5. Save the plan, select and announce the execution route, and proceed unless authorization or safety boundary requires user intervention.
• The plan must answer: problem, baseline, files, compat, verification, risks, retirement. If the slice adds a new owner/contract/schema/public API/architecture boundary/etc., escalate to a durable plan.
• Announce at start: "I'm using the writing-plans skill to create the implementation plan." and manage execution context and input as specified.
• Use compact artifacts for checks and readiness (BaselineUsageDraft, Requirement Ready Check, Change Necessity, Existence Check, Architecture Integrity Lens, Plan Pressure Test, Execution Readiness View, Plan-Time Complexity Check) prior to task decomposition.
• Save plans to the plans/ path and, if needed, initialize or append to docs/aegis workspace according to the rules.

When to use it

• Use when you have approved requirements for a new workstream or escalation trigger and you need an executable plan before coding.
• Use Planless Slice Lane for micro-slices within an existing parent plan/spec.
• Escalation from Planless Slice Lane to a durable plan when the slice touches new ownership, schema, public API, migration, or security/permission concerns.

What it can touch

• The plan creation process references: file maps, exact file paths, exact commands, and exact outputs. The skill prescribes what files to touch and what tasks to perform, and it uses the concept of an execution route and saving the plan to a designated location. It also relies on Aegis workspace tooling when available for plan saving and index updates.

Caveats

• The approach requires explicit approval or boundary conditions for durable plans. It relies on structured checks (BaselineUsageDraft, Requirement Ready Check, Change Necessity, Existence Check, Architecture Integrity Lens, Plan Pressure Test, Execution Readiness View, Plan-Time Complexity Check). If any of these checks indicate risk or misalignment, task decomposition should be adjusted accordingly. The rules prohibit creating new plans without satisfying these checks when applicable, and require the Slice Card path for micro-slices when appropriate.