Work Item Writer

SkillDev tools

Write durable work items with scoped descriptions, testable categorized acceptance criteria, governing references, and risk-matched guards

Available today. Use it from your connected AI after setup.

Connect ahel once, and every AI you use reads what you have installed.

Then ask your AI: use the Work Item Writer skill

What this skill tells your AI

The instructions your AI receives, as published by govctl-org/govctl in .claude/skills/wi-writer/SKILL.md and read by ahel’s review.

Define durable execution scope and completion evidence. Work Items are operational records, not normative specifications, design decisions, or round-by-round journals.

This helper owns Work Item field quality. The invoking workflow owns code, lifecycle transitions, loop execution, and VCS operations.

Discovery

Before creating or editing an item, inspect existing work and reusable guards:

govctl work list active
govctl work list queue
govctl work show <WI-ID>
govctl guard list
govctl loop list open

Use govctl work --help, govctl guard --help, and subcommand help for current syntax and nested-field editing.

Hard Stops

  • Do not introduce product obligations in Work Item fields. New behavior, validation, storage, compatibility, or lifecycle rules need governing RFC authority.
  • Do not use an ADR-like explanation as a substitute for recording a design choice in an ADR.
  • Do not put progress, commands run, validation output, review status, temporary blockers, hypotheses, or next actions in description or notes.
  • Do not split a coherent outcome into Work Items for mechanical substeps.
  • Do not permanently delete a Work Item without explicit user authorization; hand the destructive operation to the invoking workflow.
  • Stop when the task lacks the governing authority needed to define its acceptance criteria.

Field Policy

Description

State what the task will accomplish, why it is needed, and the relevant scope in one concise paragraph. It describes the execution target without restating normative contract language or tracking progress.

Acceptance Criteria

Each criterion must be independently testable and use a changelog category: add, fix, change, remove, deprecate, security, or chore.

Criteria describe observable task outcomes. They should be specific enough to decide done/not-done without prescribing incidental private structure. Use chore for internal validation or documentation outcomes that should not enter the release changelog.

Correct an existing criterion with govctl work edit <ID> "acceptance_criteria[N]" --set <value>. A recognized category prefix updates both text and category; other input updates only text. The operation preserves checklist status. Use the .text child path when a recognized prefix must remain literal text, and use --tick for status changes. Quote every path containing brackets. Use --stdin for rich text containing backticks, $(), or other shell syntax.

Notes

Use notes sparingly for closure-worthy constraints, durable implementation facts, or reasons an approach should not be retried. A note should remain useful after the item is done. A Work Item moved to cancelled records its cancellation reason here. Notes cannot override an RFC or accepted ADR.

Transient execution state belongs in loop rounds or the final response. Acceptance progress belongs in criterion status.

References And Dependencies

Reference RFCs that authorize behavior and ADRs that constrain the approach. Use depends_on only for a hard execution ordering dependency. Informational relationships belong in refs.

Create separate Work Items only for independently meaningful, reviewable outcomes. Keep helper extraction, fixtures, file moves, formatting, and similar mechanical work inside the parent item. Use a multi-item loop only when the batch contains multiple durable outcomes.

Verification Policy

Select guards from the changed surface, governing references, and acceptance criteria:

  • Project default_guards are the intersection of checks needed by every Work Item.
  • Work Item required_guards add reusable checks for this task's risk domains.
  • Prefer the narrowest guard that proves the relevant behavior.
  • Use full test or lint suites for cross-cutting changes or when narrower checks cannot cover the blast radius.
  • Run one-off diagnostic commands directly rather than turning them into completion guards.

Do not duplicate an effective guard as a plain command-success criterion merely to run it twice. A concise chore criterion may summarize validation outcomes that are not fully represented by guards.

Use guard-writer when a stable reusable check is missing.

Quality Tests

A Work Item is well formed when:

  • its title is concise and action-oriented;
  • its description is task scope rather than contract, rationale, or progress;
  • every criterion is categorized, specific, and testable;
  • references provide authority for user-visible behavior;
  • dependencies represent real execution ordering;
  • notes contain only durable post-closure context;
  • selected guards match the item's risk without broad default-suite inflation; and
  • the item represents one durable outcome rather than a mechanical fragment.

Completion Evidence

Tick criteria only when their outcomes exist. The invoking workflow may move the item to done after all criteria are complete and effective guards pass. Do not manually rerun those same guards immediately before the transition.

Run govctl check after substantive artifact edits. Use loop state for non-trivial execution memory and the commit skill for VCS operations.

Signals

GitHub stars
267
Forks
16
Last commit
Sep 2026
Advanced
Catalog kind
skill
Gateway key
wi-writer
Source
github.com/govctl-org/govctl