Back to Skills

evolve-agent

pjt222
Updated Yesterday
3 views
17
2
17
View on GitHub
Metaaidesigndata

About

The `evolve-agent` skill updates an existing agent's definition by refining its persona or creating an advanced variant. It handles assessing the agent, applying changes to its skills and tools, and updating metadata and registry references. Use it when an agent's capabilities are outdated, feedback reveals gaps, or its scope needs sharpening after real-world use.

Quick Install

Claude Code

Recommended
Primary
npx skills add pjt222/agent-almanac -a claude-code
Plugin CommandAlternative
/plugin add https://github.com/pjt222/agent-almanac
Git CloneAlternative
git clone https://github.com/pjt222/agent-almanac.git ~/.claude/skills/evolve-agent

Copy and paste this command in Claude Code to install this skill

Documentation

演既有員

改、擴、或造以 create-agent 所造員之進變體。此術司員生之維:按最佳實踐評缺、施針對改於人格定、升版、保持 registry 與互引同步。

用時

  • 庫新技後員之技列過時
  • 用戶回示缺能、目不明、例弱
  • 工具需易(新 MCP 服、工具去、權減)
  • 員目須銳——與他員重或過廣
  • 原員外需進變體(如 r-developerr-developer-advanced
  • 相關員或團新立,See Also 之互引陳

  • 必要:欲演員文之路(如 agents/r-developer.md
  • 必要:演之因(回、新技、工具易、目重、團合、察限)
  • 可選:版升之大(patch、minor、major)
  • 可選:是否造進變體而非就地改(默:就地改)

第一步:評當員

讀員文並按 guides/agent-best-practices.md 之質清單評各節:

SectionWhat to CheckCommon Issues
FrontmatterAll required fields present (name, description, tools, model, version, author)Missing tags, stale version, wrong priority
PurposeSpecific problem statement, not generic "helps with X"Vague or overlapping with another agent
CapabilitiesConcrete, verifiable capabilities with bold lead-insGeneric ("handles development"), no grouping
Available SkillsMatches frontmatter skills list, all IDs exist in registryStale IDs, missing new skills, lists default skills unnecessarily
Usage Scenarios2-3 realistic scenarios with invocation patternsPlaceholder text, unrealistic examples
ExamplesShows user request and agent behaviorMissing or trivial examples
Limitations3-5 honest constraintsToo few, too vague, or missing entirely
See AlsoValid cross-references to agents, guides, teamsStale links to renamed or removed files
# Read the agent file
cat agents/<agent-name>.md

# Check frontmatter parses
head -20 agents/<agent-name>.md

# Verify skills in frontmatter exist in registry
grep "skills:" -A 20 agents/<agent-name>.md

# Check if agent is referenced by any team
grep -r "<agent-name>" teams/*.md

得: 按節組之具體缺、弱、或改機之列。

敗則: 若員文不存或無 frontmatter,此技不適——當用 create-agent 從頭造。

第二步:集演之求

識並類演之因:

TriggerExampleTypical Scope
User feedback"Agent missed XSS in review"Add skill or capability
New skills availableLibrary gained analyze-api-securityUpdate skills list
Tool changeNew MCP server availableAdd to tools/mcp_servers
Scope overlapTwo agents both claim "code review"Sharpen purpose and limitations
Team integrationAgent added to a new teamUpdate See Also, verify capabilities
Model upgradeTask requires deeper reasoningChange model field
Privilege reductionAgent has Bash but only reads filesRemove unnecessary tools

編前記所需具體改。各改附目節:

- Frontmatter: add `new-skill-id` to skills list
- Capabilities: add "API Security Analysis" capability
- Available Skills: add `new-skill-id` with description
- Limitations: remove outdated limitation about missing skill
- See Also: add link to new team that includes this agent

得: 具體改之列,各映員文某節。

敗則: 若改不明,進前請用戶澄。模糊之演目生模糊之改。

第三步:擇演之範

用此決表定就地改或造變體:

CriteriaRefinement (in-place)Advanced Variant (new agent)
Agent IDUnchangedNew ID: <agent>-advanced or <agent>-<specialty>
File pathSame .md fileNew file in agents/
Version bumpPatch or minorStarts at 1.0.0
ModelMay changeOften higher (e.g., sonnet → opus)
RegistryUpdate existing entryNew entry added
Original agentModified directlyLeft intact, gains See Also cross-reference

就地改:更技、修文、銳目、調工具時。員保其身。

變體:演後版將事顯異之眾、需異模、或加能使原過廣時擇。原留以簡用。

得: 明決——改或變——附理。

敗則: 不確時默改。後可析變體;合回更難。

第四步:施改於員文

於改

直編員文:

  • Frontmatter:依需更 skillstoolstagsmodelprioritymcp_servers
  • Purpose/Capabilities:改以映新範或加能
  • Available Skills:加新技附述,去廢
  • Usage Scenarios:加或改場景以示新能
  • Limitations:去不適之限,加新誠者
  • See Also:更互引以映當前員/團/指之景

循此編則:

  • 保諸有節——加容,勿去節
  • Available Skills 節與 frontmatter skills 列同步
  • 勿加默技(meditateheal)於 frontmatter,除非為員方法之核
  • 驗各技 ID 存:grep "id: skill-name" skills/_registry.yml

於變體

# Copy the original as a starting point
cp agents/<agent-name>.md agents/<agent-name>-advanced.md

# Edit the variant:
# - Change `name` to `<agent-name>-advanced`
# - Update `description` to reflect the advanced scope
# - Raise `model` if needed (e.g., sonnet → opus)
# - Reset `version` to "1.0.0"
# - Expand skills, capabilities, and examples for the advanced use case
# - Reference the original in See Also as a simpler alternative

得: 員文(改或新變體)過第一步之評清單。

敗則: 若編破文構,用 git diff 審改並以 git checkout -- <file> 回退部分編。

第四·五步:同譯變體

譯存時必需。 此步適人與 AI 二者。勿略——陳 source_commitnpm run validate:translations 於諸 locale 生誤陳警。

察演員是否有譯並更譯以映新源態:

# Check for existing translations
ls i18n/*/agents/<agent-name>.md 2>/dev/null

譯存之時

  1. 取當源提交哈:
SOURCE_COMMIT=$(git rev-parse HEAD)
  1. 於各譯文 frontmatter 更 source_commit
for locale_file in i18n/*/agents/<agent-name>.md; do
  sed -i "s/^source_commit: .*/source_commit: $SOURCE_COMMIT/" "$locale_file"
done
  1. 於提交訊中標所涉 locale 以供重譯:
evolve(<agent-name>): <description of changes>

Translations flagged for re-sync: de, zh-CN, ja, es
Changed sections: <list sections that changed>
  1. 重生譯狀文:
npm run translation:status

無譯之時

無須行。進第五步。

於變體

新變體之譯延至變體穩(1-2 版)。變體改至少一次後方加譯。

得: 諸譯文 source_commit 更至當提交。npm run translation:status 出 0。

敗則:sed 不配 frontmatter 域,手開譯文驗其 YAML frontmatter 含 source_commit。若缺,以 npm run translate:scaffold -- agents <agent-name> <locale> 再造。

第五步:更版與元數

於 frontmatter 升 version,循語義版:

Change TypeVersion BumpExample
Typo fix, wording clarificationPatch: 1.0.0 → 1.0.1Fixed unclear limitation
New skills added, capability expandedMinor: 1.0.0 → 1.1.0Added 3 new skills from library
Restructured purpose, changed modelMajor: 1.0.0 → 2.0.0Narrowed scope, upgraded to opus

亦更:

  • updated 為今日
  • tags 若域覆易
  • description 若目實異
  • priority 若員對他之重易

得: Frontmatter versionupdated 映改之大與日。新變體始於 "1.0.0"

敗則: 若忘升版,下演無法辨當態與前。提交前必升。

第六步:更 registry 與互引

於改

agents/_registry.yml 中既條以合改後 frontmatter:

# Find the agent's registry entry
grep -A 10 "id: <agent-name>" agents/_registry.yml

descriptiontagstoolsskills 以合員文。無須改計。

若員能或名易,更他文之互引:

# Check if any team references this agent
grep -r "<agent-name>" teams/*.md

# Check if any guide references this agent
grep -r "<agent-name>" guides/*.md

於變體

按字母位於 agents/_registry.yml 加新員:

  - id: <agent-name>-advanced
    path: agents/<agent-name>-advanced.md
    description: One-line description of the advanced variant
    tags: [domain, specialty, advanced]
    priority: normal
    tools: [Read, Write, Edit, Bash, Grep, Glob]
    skills:
      - skill-id-one
      - skill-id-two

然後:

  1. 增 registry 頂之 total_agents
  2. 於原員 See Also 加指變體之引
  3. 於變體 See Also 加指原之引
  4. .claude/agents/agents/ 之 symlink 令變體自動可發現

得: Registry 條合員文 frontmatter。變體時 total_agents 等實條數。

敗則:grep -c "^ - id:" agents/_registry.yml 計條數驗與 total_agents 合。

第七步:驗演員

行全驗清單:

  • 員文存於預路
  • YAML frontmatter 解無誤
  • version 已升(改)或設「1.0.0」(變)
  • updated 為今日
  • 諸必節存:Purpose、Capabilities、Available Skills、Usage Scenarios、Examples、Limitations、See Also
  • Frontmatter 之技合 Available Skills 節
  • 諸技 ID 存於 skills/_registry.yml
  • 默技(meditateheal)不列除非方法之核
  • 工具列循最小權則
  • Registry 條存且合 frontmatter
  • 變體時:total_agents 合盤實數
  • 互引雙向(原 ↔ 變)
  • git diff 示原容無誤刪
# Verify frontmatter
head -20 agents/<agent-name>.md

# Check skills exist
for skill in skill-a skill-b; do
  grep "id: $skill" skills/_registry.yml
done

# Count agents on disk vs registry
ls agents/*.md | grep -v template | wc -l
grep total_agents agents/_registry.yml

# Review all changes
git diff

得: 清單皆過。演員可提交。

敗則: 各敗項各處。最常演後問題為 Available Skills 之陳技 ID 與忘更 updated

  • 員文存且 YAML frontmatter 有效
  • version 域映所改
  • updated 日為當
  • 諸節存而內一
  • Frontmatter skills 合 Available Skills 節
  • 諸技 ID 存於 skills/_registry.yml
  • 默技不必要列
  • Registry 條合員文
  • 變體時:agents/_registry.yml 附新條附正路
  • 變體時:total_agents 已更
  • 互引有效(See Also 無破)
  • 有譯之改:諸 locale 文 source_commit 已更
  • git diff 確無誤刪

  • 忘升版:無升則不能追何變何時。提交前必更 frontmatter versionupdated
  • 演後陳譯:庫中 1,288+ 譯文,每員演引至 4 locale 文陳。必以 ls i18n/*/agents/<agent-name>.md 察並更各 source_commit,或於提交訊標供重譯
  • 技列漂:Frontmatter skills## Available Skills 節必同步。一更而不更另致人機皆惑
  • 不必列默技:於 frontmatter 加 meditateheal,彼已自 registry 承。唯為方法之核時列(如 mysticalchemist
  • 演時工具過給:演時加 BashWebFetch「以備」。每工具加須具體新能證
  • 變體後 See Also 陳:造變體時,原與變須互引。單向引令圖不全
  • Registry 條未更:員技、工、述改後必更 agents/_registry.yml。陳條致發現與工具敗

  • create-agent — 造新員之基;evolve-agent 假此先循
  • evolve-skill — 並行演 SKILL.md 文之術
  • commit-changes — 以述訊提交演員

GitHub Repository

pjt222/agent-almanac
Path: i18n/wenyan/skills/evolve-agent
0
agentsagentskillsai-assisted-developmentclaude-codeskillsteams

Related Skills

content-collections

Meta

This skill provides a production-tested setup for Content Collections, a TypeScript-first tool that transforms Markdown/MDX files into type-safe data collections with Zod validation. Use it when building blogs, documentation sites, or content-heavy Vite + React applications to ensure type safety and automatic content validation. It covers everything from Vite plugin configuration and MDX compilation to deployment optimization and schema validation.

View skill

polymarket

Meta

This skill enables developers to build applications with the Polymarket prediction markets platform, including API integration for trading and market data. It also provides real-time data streaming via WebSocket to monitor live trades and market activity. Use it for implementing trading strategies or creating tools that process live market updates.

View skill

creating-opencode-plugins

Meta

This skill helps developers create OpenCode plugins that hook into 25+ event types like commands, files, and LSP operations. It provides the plugin structure, event API specifications, and implementation patterns for JavaScript/TypeScript modules. Use it when you need to intercept, monitor, or extend the OpenCode AI assistant's lifecycle with custom event-driven logic.

View skill

sglang

Meta

SGLang is a high-performance LLM serving framework that specializes in fast, structured generation for JSON, regex, and agentic workflows using its RadixAttention prefix caching. It delivers significantly faster inference, especially for tasks with repeated prefixes, making it ideal for complex, structured outputs and multi-turn conversations. Choose SGLang over alternatives like vLLM when you need constrained decoding or are building applications with extensive prefix sharing.

View skill