agentleFS
axiom-desuaxiom-desu/ZCodium/.agents/skills/feature-boundary-planner/SKILL.md

feature-boundary-planner

Skill37 starsChanged 57 days agoApache-2.0View on GitHub

Map a ZCode behavior change to current UI surfaces, state owners, protocol commands, persistence, and validation. Use for impact analysis, product-boundary planning, or an implementation handoff grounded in the checked-out source.

5 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use e2e-case-lifecycle, zcode-large-task-fixture and 1 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • ZCodium-projectSkill

    e2e-case-lifecycle

    Plan, maintain, replay, and promote ZCode desktop E2E cases and their fixtures, coverage matrices, and CI admission evidence.

    Also in 3 projects that use this

  • ZCodium-projectSkill

    zcode-large-task-fixture

    Generate, verify, or remove large synthetic ZCode task fixtures in local ~/.zcode persistence for UI/session performance testing. Use when testing big conversation/task rendering, task list performance, snapshot recovery, tool-call rendering, reasoning, patches, todos, and model usage without spending time chatting manually.

    Also in 3 projects that use this

  • WillhongSkill

    session-management

    Also in 3 projects that use this

The file

We split this skill into the rules it gives an agent

8 rules. 0 are common, 6 are distinctive, and 2 are very specific to this project.

  • 0

    Common

    Most of the 4 edited versions we read keep these.

  • 6

    Distinctive

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Use the mode implied by the request.

    line 16
  2. Very specific:

    Keep existing node IDs for unchanged semantic boundaries and add aliases for new terminology.

    line 36
  3. Distinctive:

    Before handing off validation, check which tests, fixtures and commands actually exist in the target package.

    line 48
  4. Distinctive:

    Use impact-brief-template.md for the relevant parts of the result:

    line 52
  5. Very specific:

    Keep desktop desktop-continuous delivery separate from mobile web-remote-replayable recovery.

    line 64
  6. Distinctive:

    Treat runtime state, snapshots, indexes, settings and caches as distinct until their synchronization is proven.

    line 66
  7. Distinctive:

    Do not infer current functionality from a directory left behind by build artifacts, an old document, or a historical branch.

    line 67
  8. Distinctive:

    Do not silently change product semantics to match an implementation discrepancy.

    line 68

“Common” is measured: kept by at least 6 in 10 of the 4 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Feature Boundary Planner

Trace the requested behavior through the current checkout. A feature belongs in the result only when current source or an explicit new requirement supports it. Verify referenced paths and symbols before using them; do not reconstruct removed functionality from historical catalogs.

Choose The Scope

  • impact-only: inspect and report without changing code or product specs.
  • planning: establish behavior and acceptance cases before implementation.
  • implementation-handoff: turn confirmed decisions into a bounded implementation and validation plan.

Use the mode implied by the request. Clarify only unknown decisions that materially change the scope; do not ask the user to reconfirm facts already established in this task.

Find Current Evidence

  1. Search aliases and node IDs in zcode-feature-graph.yaml for the user's terms. Read only matched nodes and their one-hop relationships, then verify the declared files, symbols and semantics against the current checkout. The graph is a curated seed index, not a complete feature inventory. Use source-discovery.md to fill gaps or start when there is no match. Read the relevant existing contracts and package scripts; read DESIGN.md for UI work and CONTEXT.md for plugin-store work.
  2. Locate the entrypoint with rg --files and focused rg -n searches. Trace direct callers with pnpm dep:refs <file>:<symbol> when the symbol is a TypeScript export. If an indexed codegraph tool is available, use it as additional evidence and verify its paths against the checkout.
  3. Classify the change: presentation, option source, draft/default, validation, commit effect, persistence, or recovery.
  4. Trace each user surface separately through validation and the command that commits the change. Shared UI components do not establish shared state or side effects.
  5. Identify the authoritative owner, derived views, protocol boundary, persistence and failure behavior. Name the semantic reason for each upstream or downstream dependency; imports alone do not prove a product relationship.
  6. Inspect one meaningful semantic hop first, expanding only when an unresolved owner or caller requires it. Rank relationships as must-inspect, should-inspect, conditional, invariant-only, or evidence-only.

For stateful or asynchronous behavior, show a concise diagram:

user action → surface draft → validation → owner command → event / persistence
                                                └── derived UI projection

Maintain The Seed Graph

Keep existing node IDs for unchanged semantic boundaries and add aliases for new terminology. Report missing seeds or changed relationships as graph-drift-candidate; static reachability alone does not establish a product dependency.

In impact-only mode, report proposed graph changes without editing files. In planning or implementation, update only verified entries within the task's scope. Follow the graph contract: check YAML parsing, unique IDs, relationship endpoints and ranks, and tracked source paths and symbols. Do not restore missing historical docs or claim test coverage from a graph entry.

Plan And Prune

In planning modes, write or update a feature spec before implementation. Create the spec directory if needed; do not assume an existing case catalog, test runner, or coverage workflow.

Select only dimensions that can change behavior: state, event, target, client mode, delivery kind, workspace identity, runtime availability and persistence source. Prefer representative and high-risk combinations over a global Cartesian product.

Classify cases as accepted, undefined, pruned, ignored, or bug-candidate. Give every pruned case an invariant or guard; leave undefined behavior as a concrete decision. Record accepted cases with setup, action, assertions and required evidence using case-planning-template.md.

Before handing off validation, check which tests, fixtures and commands actually exist in the target package. Distinguish planned tests, executed tests and admitted regression coverage. A missing test path is a gap, not coverage.

Output

Use impact-brief-template.md for the relevant parts of the result:

  • behavior summary and scope;
  • UI surface matrix, shared implementation and divergent behavior;
  • ranked dependencies with current file/symbol evidence;
  • state owners, validation points, commit commands and persistence;
  • invariants and representative validation cases;
  • unresolved decisions, unavailable evidence and relevant graph drift or updates.

Boundaries

  • Preserve workspaceIdentity?.trim() || workspacePath for isolation and workspacePath for execution/display.
  • Keep desktop desktop-continuous delivery separate from mobile web-remote-replayable recovery.
  • Trace accepted commands to their authoritative owner; do not turn a client draft or optimistic overlay into another accepted queue.
  • Treat runtime state, snapshots, indexes, settings and caches as distinct until their synchronization is proven.
  • Do not infer current functionality from a directory left behind by build artifacts, an old document, or a historical branch.
  • Do not silently change product semantics to match an implementation discrepancy; report the evidence and the decision needed.

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.