---

name: skill-evolver description: >- Researches and improves a specified existing Skill using successful/failed work and competitor evidence. Use for 优化这个Skill, 从成功和失败经验改进Skill, or 研究竞品后优化指定技能. Routine creation and editing use skill-creator. argument-hint: "<target Skill> [desired result or evidence]"

Daymade Skill Evolver

Orchestrate a research-driven improvement round for a named existing Skill through the current specialist Skills. Keep the original result, stage inputs, dependencies and returned evidence bound together. An unchanged Skill is a valid outcome when the evidence supports keeping it.

Keep orchestration here; execute specialist work through its current owner. Skill Creator makes and validates selected changes; it does not choose this round's business goal or research plan. When the request is diagnosis-only, remain read-only.

Resolve the target and result

Resolve the named Skill to its authoritative repository or project source, current release identity, and actual loaded entry when relevant. Read its entry and the resources used by the affected jobs. Distinguish source, installed copies and host consumption; do not edit a cache as if it were the owner.

State the user-visible result, the observation that could disprove it, supported jobs to preserve, authorized deliverables, and stopping condition before selecting improvements. Include the original reported failure. Do not replace it with a research report, a shorter prompt, a score, or green checks. Ask only for missing information or a user-owned choice that changes the work; reuse existing answers and authorization.

Keep the outcome criterion stable while comparing candidates. Correct a proxy against the original requirement before accepting an improvement; do not move the goal to whichever output the candidate already produces.

Gather evidence that can change a decision

Start with current implementation, project decisions, existing tools/Skills, provided examples, and previously verified work. Reuse evidence whose source and conditions still apply under the current owner's contract. Before executing a selected stage, load stage handoffs and follow its applicable owner route; this reference specifies inputs and returned results, not a second specialist procedure.

Select stages from the user's explicit request, accepted route and unresolved questions. A narrow repair with decisive existing evidence need not run unrelated research, but its simplicity does not waive a requested history, favorites, competitor or Deep Research lane. Mark reused, executed, unnecessary and blocked work accurately; do not describe skipped research as completed.

For each selected stage, preflight an available, capability-matched owner, its relevant branch, permissions and required data; then load and execute it. Prefer available public owners for the public workflow. Pass the original user goal, target source/ref, old successful jobs, authorization, prepared evidence and question to resolve. Execute that workflow and inspect its actual evidence or result before continuing; an owner list or a load receipt alone is not a completed stage. Keep gaps attached to the dependent decision rather than rebuilding an evidence list from the recommendation.

Use the history-reader and Skill Creator handoffs for authorized history and prepared cases, including successes, failures and user corrections. A positive opportunity does not require an earlier failure in our Skill. A request to audit a whole session and route its lessons uses the public history/Creator handoff, with session-retrospective only as an accessible optional local adapter; ordinary target improvement does not require a separate retrospective receipt.

Research the target's task capabilities

Bind research questions to the business result and target's input, error, recovery and downstream-consumer paths. Use accessible favorites-search or supplied curated links/exports through the reference's curation handoff before external tool or method selection, then execute competitors-analysis's source and synthesis routes for relevant capabilities. Return their source-bound mechanisms and counterevidence to this round; do not substitute a feature list for the owner's decision chain.

Execute deep-research for requested research lanes or unresolved breadth/synthesis, using its current asset, mode and source contracts. Reuse accepted routes and existing original exports; do not force a provider, paid mode or report length. Stop retrieval when remaining answers cannot change selection or required acceptance. Keep source-contract facts, exercised behavior and unmeasured business benefits separate when joining the returned evidence.

Decide every important mechanism

Use Skill Creator's applicable ordinary-update or cross-task-improvement handoff to classify evidence-backed findings and identify the smallest maintenance point. Use competitors-analysis's Landscape handoff to account for every important discovered mechanism as adopt, do not adopt or needs verification under its existing disposition contract. Check the returned coverage against the original questions and sources, including positive opportunities and existing equivalents; do not manufacture changes.

When a technical alternative must be chosen, execute tech-selection's applicable full or lightweight route with the joined evidence and current authorization. Carry its actual choice or unresolved user-owned decision into implementation. If an existing owner already covers the need, repair its loading or invocation instead of duplicating it. Combine compatible improvements; leave major direction, new product scope, real preferences and irreversible/external choices at their existing authorization boundary. Choose and execute necessary local reversible implementation details within the authorized target work.

Implement and verify the selected changes

Freeze the complete target bundle from an immutable source before editing. Preserve its triggers, supported jobs, interfaces, confirmation points, recovery paths and user-owned data. Work in an isolated source checkout when the shared checkout carries other work. Keep the target's source and directly affected instructions, examples, help and caller routes aligned.

Inventory affected runtime defaults and side effects before replacing a mechanism, including read/write behavior, access permissions and generated-output locations. Preserve them unless the authorized improvement explicitly changes them and its acceptance checks cover that change. Do not infer a new requirement from a general security preference.

Execute the Skill Creator making, migration and verification handoff for the selected change under the repository's normal workflow. Pass the frozen business criterion, old jobs and defaults/side effects as preservation requirements. Use the target's execution owner for its actual consumer and recovery check; do not delegate away responsibility for whether the joined result meets the request. Classify changes by their actual failure surface; new capability or methodology does not automatically authorize paired evaluation, benchmarking, repeated trials, a viewer, or extra agents.

Verify the original result through its real consumer, including the affected failure and recovery path. Check important old jobs against the frozen bundle. Separate static preservation from task-result evidence. Calibrate acceptance with known healthy and failing cases; missing inputs and empty values are distinct when the rule examines them. Source reading, a subprocess exit, tests or CI may support a claim but cannot substitute for the result being requested.

Complete authorized delivery through the target repository's release and local availability procedures. Read back the exact published or consumed object when that is part of the requested result. Keep human acceptance unconfirmed when it has not been observed; still finish the authorized implementation and checks.

Use available executors without losing the result

Resolve selected owners by capability through available skills-search; if it is absent, inspect the current host catalog and user-declared accessible sources. Evolver ships only its own bundle, not companion Skills. Public availability does not prove installation, branch readiness or access to required data. Do not require our private repositories, credentials or local configuration. Do not automatically install/clone an owner or override a host's disabled-Skill policy. For a public owner not installed, use the public catalog and installation guide to identify its actual plugin package, or use an available equivalent owner or tools; do not invent a successful invocation.

The same agent may load and execute Skills sequentially. Before authorized delegation or parallel independent lanes, follow the current user/host contract and public owner's delegation/verification workflow. Apply daymade-agent-discipline only when that optional local adapter is already present and accessible; dependent stages wait for the evidence they need. Do not assign one agent per Skill by default or copy an owner's engine here.

An absent invocation tool does not make a resolved Skill absent: read its current entry and execute the applicable workflow with available tools.

If a specialist is absent, use available tools for the same substantive work: read bounded supplied records, inspect fixed source objects, run the needed probe, compare trade-offs, or edit and validate the canonical source. Do not invent a specialist invocation or claim its gate passed. If a required source, permission, runtime or consumer is unavailable, pause only that dependent item, name the missing observation and resume condition, and continue independent authorized work. Missing optional tooling does not waive a required outcome.

Stop and leave an incremental continuation

Report the actual improvement or evidence-backed no-change decision, the exact source/diff and result evidence, and the remaining unmet criteria. Keep source, publication, installation, prepared-task runs and live business results distinct.

Update only changed decisions and source anchors in the target project's existing canonical document when one exists. Keep reusable corrections in their owning Skill; preserve raw historical records as history. Leave the next unmet action and its reopening condition in the existing task context rather than creating memory or a second state ledger. Resume from that action instead of repeating settled research. Stop when required results and delivery are verified, or when only explicitly blocked items remain; do not start an automatic improvement loop.