App Evolver
Improve the application's actual user result. Extract transferable mechanisms, resolve their fit with the current product, implement authorized improvements, and verify them through the application's consuming surface.
Use this as an orchestration Skill. Keep repository engineering, research,
history access and release operations with their existing owners. For a Skill
target use skill-evolver; for competitor intelligence alone use
competitors-analysis; for a new product specification use the available
requirements workflow. Do not replace those owners with another execution engine.
1. Bind the application and unfinished result
Before research or edits, establish:
- Exact application, repository/ref, platform and runtime or deployment identity.
- Original user request, current product contract and active stage; distinguish long-term positioning from the work selected now.
- User-visible improvement, falsifiable acceptance evidence, authorized deliverables, protected behavior/data and stopping condition.
- Analysis-only versus implementation authorization, and any unresolved product choice whose answer changes the outcome.
Reuse earlier user decisions; ask in the current conversation only for missing information or a real unresolved choice. Use the host's question tool when available. State the evidence, alternatives, consequences and recommendation. Do not delegate a user decision through another agent.
For multiple applications, bind each contract and acceptance path separately. A successful design, permission or measurement in one app is an unverified proposal in another. Do not transfer a desktop contract to its mobile sibling without checking the sibling's current contract.
Read repository instructions and relevant domain Skills. Pin the code used for judgment to an immutable ref; inspect current runtime/config separately. Preserve other writers' work and coordinate overlapping ownership before editing. On resume, reconcile the latest request, actual artifacts and unfinished evidence; continue from the first unmet result rather than replaying the original plan.
2. Ground opportunities and constraints
Read evidence and fit before selecting changes.
Inspect the existing user journey, assets, services, tests and implementation owners first. Search available prior work and personal collections when requested or required by the user's environment. Perform explicitly requested research lanes; do not silently replace deep research or history mining with a quick read. Stop a lane once its evidence resolves the decision it was meant to inform.
Use history only within explicit authorization. Retrieve through the owning index and exact session reader, preserving user/assistant/tool roles and child-agent prompt/result links. Exclude hidden reasoning. Keep unavailable, stale or incomplete coverage explicit; do not fall back to an unbounded raw-history scan. Treat assistant conclusions as hypotheses, not user endorsement or current truth.
For competitors, inspect implementation and current primary evidence beyond a
feature list. Use competitors-analysis for repository evidence and the available
research owner for non-repository products. Resolve revisions and distinguish
observed behavior from claims. Treat source access, architectural inspiration,
code reuse and distribution permission as separate questions; reading a repository
does not authorize copying its implementation.
3. Decide what transfers
For each mechanism that could change the next action, record:
| Decision input | Required answer |
|---|---|
| Evidence | Exact source/ref and observed normal, failure or recovery behavior |
| User benefit | Which current user result improves, and how it can be falsified |
| Existing equivalent | What the app already provides and where it differs |
| Fit and cost | Contract, lifecycle, data, defaults, dependencies and operational cost |
| Disposition | Adopt, adapt, reject or needs verification, with the decisive reason |
| Next evidence | Smallest probe or journey that resolves remaining uncertainty |
Keep positive opportunities even when no existing defect has been demonstrated. Do not invent failures to justify them. Reject mechanisms that contradict the current contract; preserve useful parts without importing the competitor's whole architecture, business model or weaker guarantees. A failed parameter family is not a reason to try indefinitely; identify what structurally new evidence would justify another attempt.
When the user requests a team, use the host's actual collaboration tools and bounded roles: product fit, implementation/lifecycle and independent challenge as needed. Give original requirements, exact evidence, distinct ownership and falsifiable probes. Treat findings as hypotheses until the executor checks their sources or runs the probes. Without such authorization, use the minimum required independent review and do not start a research or evaluation fan-out.
Resolve researchable uncertainty yourself. Ask the user about preferences, mutually exclusive product choices or missing authorization, not facts available in code or a read-only probe. A missing optional research lane does not block independent authorized work; a missing load-bearing contract blocks its dependent change until resolved.
4. Implement the smallest useful increment
Read implementation and acceptance before editing or launching validation.
Route implementation to the repository's actual engineering workflow and domain owner, not Skill Creator. If an optional Skill is unavailable, use accessible repository instructions and applicable tools within authorization; do not auto-install private dependencies or pretend their checks ran.
Order work by user benefit and dependencies. Preserve protected behavior, source provenance, persistence, identity and recovery paths. Keep optional work out of a mandatory hot path when the product contract requires independent completion; choose its limits from this app's workload and measurements. Do not copy a timeout or device policy from an unrelated app.
Complete requested implementation and necessary verification; do not stop at a plan or recommendation when the user asked to optimize the application. Keep an analysis-only request read-only. Respect the authorized boundary for deployment, production writes, spending, publication and destructive operations.
5. Verify, correct and deliver
Verify the intended consumer's outcome and protected healthy cases. Separate source checks, synthetic tests, real runtime journeys, packaged/installed behavior, production state and human adoption. Evidence at one layer does not establish another. Read back at the point where the result is consumed.
When evidence fails, inspect raw signals, seek a falsifier and correct the root cause or the mistaken criterion. Retry only after a relevant input or hypothesis changes. Calibrate new guards/checks on healthy and failing inputs; account for examined items and demonstrate that the intended bad behavior is detected. Do not weaken a product guarantee merely to turn a check green.
Finish repository delivery through its established authorized workflow. Update only changed facts in existing product SSOTs. Keep reusable lessons with their owning Skills when requested; do not create memory files, hooks or new processes as an automatic by-product.
Report what changed, why it fits, where it landed, what user-result evidence passed and what remains unverified. Request human acceptance when the declared outcome depends on the user's judgment; never mark it passed on their behalf. Stop when the authorized result and necessary evidence are complete. If only an external condition remains, name the unfinished result and exact resume condition.
