modular-evolution
/modular-evolution
Keep software evolution modular and maintainable while controlling duplicated candidates, dependencies, environments, and artifacts. Use for modularization and consolidation decisions, candidate or worktree lifecycle, rebuildability, dependency and disk inventories, or evidence-preserving retirement. Do not trigger for every small code edit, generic performance tuning, or unrelated file cleanup.
- Author
- phenomenoner ↗
- Repository
- modular-evolution ↗
- Licence
- MIT
- Last repository change
- 07.11.2026
Assessment and security
- Files in this package
- 7
- Scripts
- 0
- HTML comments
- 0
- URLs in SKILL.md
- 0
No signals detected. Repository changed on 07.11.2026, without references to retired models or APIs.
MIT, open: use and modify while retaining author attribution.
What it does and when to use it
Keep software evolution modular and maintainable while controlling duplicated candidates, dependencies, environments, and artifacts. Use for modularization and consolidation decisions, candidate or worktree lifecycle, rebuildability, dependency and disk inventories, or evidence-preserving retirement. Do not trigger for every small code edit, generic performance tuning, or unrelated file cleanup.
Original author description
Keep software evolution modular and maintainable while controlling duplicated candidates, dependencies, environments, and artifacts. Use for modularization and consolidation decisions, candidate or worktree lifecycle, rebuildability, dependency and disk inventories, or evidence-preserving retirement. Do not trigger for every small code edit, generic performance tuning, or unrelated file cleanup.
How it works
- Scope and authority
- Prevent unnecessary growth during development
- Practical workflow
- Quick decisions
SKILL.md preview
--- name: modular-evolution description: "Keep software evolution modular and maintainable while controlling duplicated candidates, dependencies, environments, and artifacts. Use for modularization and consolidation decisions, candidate or worktree lifecycle, rebuildability, dependency and disk inventories, or evidence-preserving retirement. Do not trigger for every small code edit, generic performance tuning, or unrelated file cleanup." metadata: version: "0.2.0" --- # ModE (Modular Evolution) > Force your agent to think modularly and keep the best of your software’s evolution. Keep useful differences, rebuildability, and evidence while controlling the growth of candidates, environments, and process. Judge success by the user's actual workflow. Saving storage must preserve recovery capability. Before adding a module, full copy, environment, or candidate, consider direct reuse and explicit compatibility boundaries. The best results include validated behavior, maintainable source, counterexamples, necessary history, and recovery data. Read [product decisions](references/product-decisions.md) only when maintaining positioning or comparing approaches. ## Scope and authority Choose the requested activity: inventory, planning, implementation, or retirement. Do not automatically chain every phase. A review or plan does not authorize refactoring, restarting paused systems, moving data, deleting files, committing, or publishing. Complete work already covered by the user's authority; ask only for a material unresolved decision or an effect outside that authority. Follow applicable system, developer, user, host, and project rules. This skill supplies a method, not a replacement for retention or confirmation policies. When selecting scratch, archive, or installation locations, read [host integration](references/local-host.md) and the actual local instructions. Preserve compatible safeguards without inventing repeated approval gates. ## Prevent unnecessary growth during development - Keep one maintained canonical source. Use commits, differences, and worktrees when concurrent work needs them instead of copying the entire project for every experiment. Worktrees share Git objects; checked-out files, dependencies, and output still cost space. - Split modules along demonstrated responsibility, change, and execution boundaries. Make inputs, outputs, ownership, and compatibility explicit. Prefer an existing module, workspace, interface, or configuration over new services, registries, or release infrastructure. - Move shared code out of candidate-specific paths when coupling justifies it. Use explicit modules, pinned packages, or managed artifact references; migrate and verify consumers before retiring the old path. - Reuse tool-managed downloads or content-addressed stores while preserving required environment isolation. Matching lockfiles do not prove matching OS, CPU, ABI, toolchain, build flags, or install side effects. - Separate rebuildable output from unique evidence and private state. Credentials, profiles, databases, recovery checkpoints, unbacked-up data, and crash reproducers remain valuable regardless of folder names. Consult [toolchains and Windows](references/toolchains-and-windows.md) when selecting a reuse mechanism or migration method. ## Practical workflow ### 1. Establish the outcome and current state Identify whether the task prevents future growth, consolidates selected branches, or retires known output. Fix the writable scope and systems that must stay paused. Reuse existing indexes, then perform a bounded read-only inventory of selected roots. Record source and uncommitted changes, locked inputs, outputs, evidence, private state, consumers, cross-candidate references, and storage. Start with the largest relevant categories. Distinguish logical file size, unique file objects, allocated storage, and actual free capacity. Report inaccessible paths, links, and incomplete scans as gaps. Use [inventory and retirement](references/inventory-and-retirement.md) for the detailed decision criteria. ### 2. Choose the smallest reversible unit Pick one concrete source of growth: a repeated dependency environment, a cross-version module reference, or a class of rebuildable output. State the intended reduction, compatibility boundary, verification, and recovery path. Extract a module only when repeated changes, mixed responsibilities, or coupling supply evidence; do not make a whole-system rewrite a cleanup prerequisite. If the structure already works, improve pinned inputs, explicit references, or the copying practice directly. Set environment counts or retention budgets from measured needs. Do not silently add monitoring, schedules, or broader retention policy. ### 3. Implement and verify within authority Preserve uncommitted differences and necessary state. Change the selected source, dependency, or build flow, then rebuild from pinned, obtainable inputs with isolated state. The rebuild must not secretly depend on a path being retired through PATH, working directory, editable installs, or global packages. Use the smallest check capable of falsifying the changed claim. Add native-host, integration, or recovery checks only when the affected behavior requires them. Do not automatically run full suites or create review chains for ordinary consolidation. Before entering repeated rebuild, retry, or review loops, set a numeric budget and stop condition; a new candidate name does not reset the budget. ### 4. Establish supersession by scope Newer is not automatically better for every purpose. Compare the old candidate's actual supported behavior with evidence for its successor. Inspect entry points, scripts, other candidates, frozen reviews, recovery procedures, and active consumers. A partial replacement leaves other purposes and evidence intact. Retirement requires replacement evidence for the selected scope, recoverable source and inputs, separately retained necessary evidence and private state, migrated consumers, no known active dependency, and applicable authority. If a condition is unknown, retain that portion and continue independent verified work. Use existing records for `active → superseded-for-scope → retireable → archived`. An `evidence-pinned` flag may coexist with any state. A small table or issue is enough; do not mandate a registry, manifest, ledger, or database. ### 5. Retire recoverably and report facts Recheck exact absolute paths, links, destinations, active use, preservation, and authority immediately before the effect. Follow the established archive method. Hardlinks, junctions, symlinks, and shared writable environments are mechanisms to verify, not safety guarantees. Do not supply unattended recursive permanent-deletion routines. Report changes, retained items, affected consumers, executed checks, recovery paths, and remaining gaps. Distinguish estimated reclaimable bytes, archived bytes, and measured capacity released. A same-volume archive still occupies that volume. Without measurements, make no savings claim; without a runtime observation, make no claim that an app restarted or the skill activated. ## Quick decisions | Situation | Next action | | --- | --- | | Candidates differ by a few source files | Preserve differences in Git; use worktrees only for concurrent work; handle environments and evidence separately. | | A new version imports an older candidate's runtime | Extract or explicitly pin the shared module, migrate consumers, then verify before retirement. | | Lockfiles or hashes match | Treat this as an investigation clue; check compatibility, state, and consumers before choosing tool-native reuse. | | Old code is replaced but holds a unique failure | Preserve a sufficient reproducer and source mapping independently; retire only the justified portion. | | A directory is called cache or backup | Establish contents, ownership, references, and rebuild or recovery conditions; name and age prove nothing. | | The request only asks for principles or an inventory | Deliver that result within scope. |
Files in this package (7)
- .gitattributes 1 KB
- .gitignore 1 KB
- CHANGELOG.md 1 KB
- LICENSE 1 KB
- README.md 9 KB
- README.zh-TW.md 8 KB
- SKILL.md 8 KB
Versions
The database has no other skills named modular-evolution.
Database snapshot: 13.11.2026 · package version db2dcf7d00b3 (SHA-256 digest of the entire package)
Downloads in the last 30 days: 0 · ranking
Report a problem with this skill: harmful content, incorrect descriptions or an author’s removal request.
Similar skills
- 100abp-modularity
ABP Framework v10.x (10.4–10.6) modularity: AbpModule, [DependsOn], module lifecycle, plugin modules, modular monolith. Use when creating a module, defining module dependencies, or building a modular architecture in ABP. - 100abp-modularity
ABP Framework v10.x (10.4–10.6) modularity: AbpModule, [DependsOn], module lifecycle, plugin modules, modular monolith. Use when creating a module, defining module dependencies, or building a modular architecture in ABP. - 100repo-maintainer
Audit and repair repository hygiene across artifacts, dependencies, CI, docs, Git state, and code-quality signals. Use for repository maintenance, cleanup, health checks, or pre-release hardening.