writing-plans-extras

Traces deferred obligations and approval gates from specifications into implementation plans.

1|Updated Dec 13, 2025
One-click install
npx skills add https://github.com/ZK-Theory/TDL --skill writing-plans-extras-zk-theory
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: writing-plans-extras
Source: https://github.com/ZK-Theory/TDL/tree/main/.agents/skills/writing-plans-extras
Command: npx skills add https://github.com/ZK-Theory/TDL --skill writing-plans-extras-zk-theory

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? When decomposing accepted specifications into implementation plans, forward-looking obligations (deferred requirements, approval gates, retention or migration duties) often sit outside the primary interface sections and silently disappear from task breakdowns. This Skill ensures every such obligation is inventoried, traced to a plan task or an explicit justified deferral, and verified before the plan is treated as executable. ## Core Features & Use Cases - Forward-Obligation Scan: Searches full governing sources (specifications, amendments, decision records) for forward-looking language and records each hit in an obligation register with source, owner, trigger, and disposition. - Authority Boundary Check: Verifies that every gate ID and dispatchable stage in a plan resolves to an accepted decision-register entry, failing plans that mint authority from proposed or reviewer-only records. - Interface Verification: Requires every named seam (function, schema, runtime call) to be confirmed against the actual file with a path:line citation, or the plan is marked interface-unverified as a dispatch blocker. - Caller-Inventory Closure: Enforces registry-aware transitive closure for "no other caller" claims, with runtime negatives for protected entries. - Use Case: Before decomposing a multi-document ARS/TDL specification into work packages, run this scan to surface deferred obligations and unresolved approval gates so the master plan closes them explicitly. ## Quick Start Use the writing-plans-extras skill to scan my accepted specification documents and draft plan for deferred obligations, unverified interfaces, and unresolved approval gates before task decomposition.

Frequently Asked Questions about writing-plans-extras

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I make sure deferred requirements are not lost when writing an implementation plan?

Run a forward-obligation scan over the full governing sources, not just summaries, searching for phrases like "deferred to", "next gate", and "requires separate approval". Record each hit in an obligation register and trace it to a numbered plan task or a named, justified deferral.

How to verify interfaces before writing plan tasks?

For every seam the plan names, open the actual file and record the real signature or behavior with a path:line citation in the plan. If you cannot verify it, mark the plan interface-unverified and treat that as an explicit dispatch blocker.

When should I use this skill alongside a plan-writing workflow?

Use it before task decomposition whenever planning from accepted specifications that may contain deferred obligations, later-gate requirements, or approval boundaries. It complements a base plan-writing skill by preventing non-local obligations from disappearing during work-package breakdown.

Can a reviewer approval close an owner gate in a plan?

No. A reviewer verdict, including an adversarial review's accept, only makes a subject eligible for a separately recorded owner decision. Every gate ID and dispatchable stage must resolve to an accepted decision-register entry, or the plan fails the authority boundary check.

Why is a direct-caller grep not enough for a "no other caller" claim?

Such claims need registry-aware transitive closure seeded from every CLI and rederivation root, following routing, coordination, and executor-registry lookups. Entries classified as protected also require a runtime negative proving capability rejection through the real root.