返回技能列表

evolve-team

pjt222
更新于 Yesterday
4 次查看
17
2
17
在 GitHub 上查看
aiautomationdesigndata

关于

The `evolve-team` skill updates an existing team composition by modifying its member roster, coordination pattern, or structure, and handles versioning and registry synchronization. Use it when a team's agents are outdated, its workflow needs refinement, or a specialized variant is required. It directly edits the team file and CONFIG block, supporting in-place updates or creating new variants.

快速安装

Claude Code

推荐
主要方式
npx skills add pjt222/agent-almanac -a claude-code
插件命令备选方式
/plugin add https://github.com/pjt222/agent-almanac
Git 克隆备选方式
git clone https://github.com/pjt222/agent-almanac.git ~/.claude/skills/evolve-team

在 Claude Code 中复制并粘贴此命令以安装该技能

技能文档

Evolve an Existing Team

Improve, restructure, or create a specialized variant of a team that was originally authored with create-team. This procedure covers the maintenance side of the team lifecycle: assessing gaps against the template and coordination patterns, applying targeted improvements to composition and workflow, bumping versions, and keeping the registry and cross-references in sync.

When to Use

  • A team's member roster is outdated after agents were added, removed, or evolved
  • User feedback reveals workflow bottlenecks, unclear handoffs, or missing perspectives
  • The coordination pattern no longer fits the team's actual workflow (e.g., hub-and-spoke should be parallel)
  • A specialized variant is needed alongside the original (e.g., r-package-review and r-package-review-security-focused)
  • Team members' responsibilities overlap and need sharper boundaries
  • The CONFIG block is out of sync with the prose description or the members list
  • A team needs to be split into two smaller teams or two teams need to be merged

Inputs

  • Required: Path to the existing team file to evolve (e.g., teams/r-package-review.md)
  • Required: Evolution trigger (feedback, new agents, coordination mismatch, scope overlap, performance issues, agent evolution)
  • Optional: Target version bump magnitude (patch, minor, major)
  • Optional: Whether to create a specialized variant instead of refining in-place (default: refine in-place)

Procedure

Step 1: Assess the Current Team

Read the existing team file and evaluate each section against the team template (teams/_template.md):

SectionWhat to CheckCommon Issues
FrontmatterAll required fields (name, description, lead, version, author, coordination, members[])Missing tags, stale version, wrong coordination
PurposeClear multi-agent justification (at least two distinct specialties)Could be handled by a single agent
Team CompositionTable matches frontmatter members, no overlapping responsibilitiesStale table, duplicated focus areas
Coordination PatternMatches actual workflow, ASCII diagram presentWrong pattern for the workflow
Task DecompositionPhased breakdown with concrete tasks per memberVague tasks, missing phases
CONFIG BlockValid YAML between markers, matches frontmatter and proseOut of sync, missing blocked_by, invalid YAML
Usage Scenarios2-3 realistic activation promptsPlaceholder text
Limitations3-5 honest constraintsMissing or too generic
See AlsoValid links to member agents, related teams, guidesStale links
# Read the team file
cat teams/<team-name>.md

# Verify all member agents still exist
grep "id:" teams/<team-name>.md | while read line; do
  agent=$(echo "$line" | grep -oP '(?<=id: )[\w-]+')
  grep "id: $agent" agents/_registry.yml || echo "MISSING: $agent"
done

# Check if the team is referenced by any guide
grep -r "<team-name>" guides/*.md

Got: A list of specific gaps, weaknesses, or improvement opportunities organized by section.

If fail: If the team file does not exist or has no frontmatter, this skill does not apply — use create-team instead to author it from scratch.

Step 2: Gather Evolution Requirements

Identify and categorize what triggered the evolution:

TriggerExampleTypical Scope
User feedback"Reviews take too long, agents duplicate effort"Sharpen responsibilities or change pattern
New agent availableapi-security-analyst agent was createdAdd member
Agent evolvedcode-reviewer gained new skillsUpdate member responsibilities
Agent removeddeprecated-agent was retiredRemove member, reassign tasks
Coordination mismatchSequential team has independent subtasksChange to parallel
Scope expansionTeam needs to cover deployment, not just reviewAdd member or create variant
Team too large6+ members causing coordination overheadSplit into two teams
Team too smallSingle member does most of the workMerge with another team or add members

Document the specific changes needed before editing:

- Frontmatter: add new member `api-security-analyst` with role "API Security Reviewer"
- Team Composition: add row to composition table
- Task Decomposition: add API security review tasks to execution phase
- CONFIG block: add member and tasks entries
- See Also: add link to new agent file

Got: A concrete list of changes, each mapped to a specific section of the team file.

If fail: If the changes are unclear, consult the user for clarification before proceeding. Vague evolution goals produce vague improvements.

Step 3: Choose Evolution Scope

Use this decision matrix to determine whether to refine in-place or create a variant:

CriteriaRefinement (in-place)Specialized Variant (new team)
Team IDUnchangedNew ID: <team>-<specialty>
File pathSame .md fileNew file in teams/
Version bumpPatch or minorStarts at 1.0.0
CoordinationMay changeMay differ from original
RegistryUpdate existing entryNew entry added
Original teamModified directlyLeft intact, gains See Also cross-reference

Refinement: Choose when adjusting members, sharpening responsibilities, fixing the CONFIG block, or changing the coordination pattern. The team keeps its identity.

Variant: Choose when the evolved version would serve a substantially different use case, require a different coordination pattern, or target a different audience. The original stays as-is for its existing use case.

Additional scope decisions:

SituationAction
Team has 6+ members and is slowSplit into two focused teams
Two teams of 2 cover adjacent domainsMerge into one team of 3-4
Team's coordination pattern is wrongRefinement — change pattern in-place
Team needs entirely different leadRefinement if lead exists; create agent first if not

Got: A clear decision — refinement, variant, split, or merge — with rationale.

If fail: If unsure, default to refinement. Splitting or merging teams has higher blast radius and should be confirmed with the user.

Step 4: Apply Changes to the Team File

For Refinements

Edit the existing team file directly. Maintain consistency across all sections that reference team composition:

  1. Frontmatter members[]: Add, remove, or update member entries (each with id, role, responsibilities)
  2. Team Composition table: Must match frontmatter members exactly
  3. Coordination Pattern: Update prose and ASCII diagram if the pattern changes
  4. Task Decomposition: Revise phases and per-member tasks to reflect new composition
  5. CONFIG block: Update members and tasks lists to match (see Step 5)
  6. Usage Scenarios: Revise if the team's activation triggers changed
  7. Limitations: Update to reflect new constraints or remove resolved ones
  8. See Also: Update agent links and add references to new related teams or guides

Follow these editing rules:

  • Preserve all existing sections — add content, do not remove sections
  • When adding a member, add them to ALL of: frontmatter, composition table, task decomposition, and CONFIG block
  • When removing a member, remove from ALL of those locations and reassign their tasks
  • Verify each member agent exists: grep "id: agent-name" agents/_registry.yml
  • Keep the lead in the members list — the lead is always a member

For Variants

# Copy the original as a starting point
cp teams/<team-name>.md teams/<team-name>-<specialty>.md

# Edit the variant:
# - Change `name` to `<team-name>-<specialty>`
# - Update `description` to reflect the specialized scope
# - Adjust `coordination` pattern if needed
# - Reset `version` to "1.0.0"
# - Modify members, tasks, and CONFIG block for the specialized use case
# - Reference the original in See Also as a general-purpose alternative

Got: The team file (refined or new variant) passes the assessment checklist from Step 1, with all sections internally consistent.

If fail: If an edit breaks internal consistency (e.g., CONFIG block lists a member not in frontmatter), compare the frontmatter members[] against the Team Composition table, Task Decomposition, and CONFIG block to find the mismatch.

Step 4.5: Sync Translated Variants

Required when translations exist. This step applies to both human authors and AI agents following this procedure. Do not skip — stale source_commit values cause npm run validate:translations to report false staleness warnings across all locales.

Check whether translations exist for the evolved team and update them to reflect the new source state:

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

If translations exist

  1. Get the current source commit hash:
SOURCE_COMMIT=$(git rev-parse HEAD)
  1. Update source_commit in each translated file's frontmatter:
for locale_file in i18n/*/teams/<team-name>.md; do
  sed -i "s/^source_commit: .*/source_commit: $SOURCE_COMMIT/" "$locale_file"
done
  1. Flag files for re-translation by including affected locales in the commit message:
evolve(<team-name>): <description of changes>

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

If no translations exist

No action needed. Proceed to Step 5.

For variants

Defer translation of new variants until the variant stabilizes (1-2 versions). Add translations after the variant has been refined at least once.

Got: All translated files have source_commit updated to the current commit. npm run translation:status exits 0.

If fail: If sed fails to match the frontmatter field, open the translated file manually and verify it has source_commit in its YAML frontmatter. If the field is missing, re-scaffold with npm run translate:scaffold -- teams <team-name> <locale>.

Step 5: Update the CONFIG Block

The CONFIG block between <!-- CONFIG:START --> and <!-- CONFIG:END --> must stay in sync with the prose sections. After any member or task change:

  1. Verify every agent in CONFIG members matches a member in the frontmatter
  2. Verify every assignee in CONFIG tasks matches a member agent id
  3. Update blocked_by dependencies if task ordering changed
  4. Ensure the synthesis/final task references all prerequisite tasks
team:
  name: <team-name>
  lead: <lead-agent>
  coordination: <pattern>
  members:
    - agent: <agent-id>
      role: <role-title>
      subagent_type: <agent-id>
  tasks:
    - name: <task-name>
      assignee: <agent-id>
      description: <one-line>
    - name: synthesize-results
      assignee: <lead-agent>
      description: Collect and synthesize all member outputs
      blocked_by: [<prior-task-names>]

Got: CONFIG YAML is valid, all agents and tasks are consistent with the rest of the file, and blocked_by forms a valid DAG.

If fail: Parse the CONFIG block YAML separately to find syntax errors. Cross-check every assignee against the members list.

Step 6: Update Version and Metadata

Bump the version field in frontmatter following semantic versioning:

Change TypeVersion BumpExample
Wording fix, See Also updatePatch: 1.0.0 → 1.0.1Fixed stale agent link
New member added, tasks revisedMinor: 1.0.0 → 1.1.0Added security-analyst member
Coordination pattern changed, team restructuredMajor: 1.0.0 → 2.0.0Changed from hub-and-spoke to parallel

Also update:

  • updated date to the current date
  • tags if the team's domain coverage changed
  • description if the team's purpose is materially different
  • coordination if the pattern changed

Got: Frontmatter version and updated reflect the magnitude and date of changes. New variants start at "1.0.0".

If fail: If you forget to bump the version, the next evolution will have no way to distinguish the current state from the previous one. Always bump before committing.

Step 7: Update Registry and Cross-References

For Refinements

Update the existing entry in teams/_registry.yml to match the revised frontmatter:

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

Update description, lead, members, and coordination fields to match the team file. No count change is needed.

For Variants

Add the new team to teams/_registry.yml:

- id: <team-name>-<specialty>
  path: <team-name>-<specialty>.md
  lead: <lead-agent>
  members: [agent-1, agent-2, agent-3]
  coordination: <pattern>
  description: One-line description of the specialized variant

Then:

  1. Increment total_teams at the top of the registry
  2. Add See Also cross-reference in the original team pointing to the variant
  3. Add See Also cross-reference in the variant pointing to the original

Run the README automation:

npm run update-readmes

Got: Registry entry matches the team file frontmatter. npm run update-readmes exits 0. For variants, total_teams equals the actual number of team entries.

If fail: If the registry count is wrong, count entries with grep -c "^ - id:" teams/_registry.yml and correct the count. If README automation fails, verify package.json exists and js-yaml is installed.

Step 8: Validate the Evolved Team

Run the full validation checklist:

  • Team file exists at the expected path
  • YAML frontmatter parses without errors
  • version was bumped (refinement) or set to "1.0.0" (variant)
  • updated date reflects today
  • All required sections present: Purpose, Team Composition, Coordination Pattern, Task Decomposition, Configuration, Usage Scenarios, Limitations, See Also
  • Frontmatter members[] matches Team Composition table
  • CONFIG block members match frontmatter members
  • CONFIG block tasks have valid assignees and blocked_by references
  • All member agent IDs exist in agents/_registry.yml
  • Lead agent appears in the members list
  • No two members share the same primary responsibility
  • Registry entry exists and matches frontmatter
  • For variants: total_teams count matches actual count on disk
  • Cross-references are bidirectional (original ↔ variant)
  • git diff shows no accidental deletions from the original content
# Verify frontmatter
head -25 teams/<team-name>.md

# Verify all member agents exist
for agent in agent-a agent-b agent-c; do
  grep "id: $agent" agents/_registry.yml
done

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

# Review all changes
git diff

Got: All checklist items pass. The evolved team is ready to commit.

If fail: Address each failing item individually. The most common post-evolution issues are CONFIG block drift (members or tasks not matching the prose) and a forgotten updated date.

Validation

  • Team file exists and has valid YAML frontmatter
  • version field reflects the changes made
  • updated date is current
  • All sections present and internally consistent
  • Frontmatter members[], Team Composition table, and CONFIG block are in sync
  • All member agent IDs exist in agents/_registry.yml
  • Lead agent is in the members list
  • CONFIG block YAML is valid and parseable
  • Registry entry matches the team file
  • For variants: new entry in teams/_registry.yml with correct path
  • For variants: total_teams count updated
  • Cross-references are valid (no broken links in See Also)
  • For refinements with translations: source_commit updated in all locale files
  • git diff confirms no accidental content removal

Pitfalls

  • CONFIG block drift: The CONFIG block, frontmatter, and prose sections must all agree on members and tasks. Updating one without the others is the most common team evolution error. After every change, cross-check all three.
  • Forgetting to bump version: Without version bumps, there is no way to track what changed or when. Always update version and updated in frontmatter before committing.
  • Stale translations after evolution: Every team evolution triggers staleness in up to 4 locale files. Always check for existing translations with ls i18n/*/teams/<team-name>.md and update source_commit in each, or flag them for re-translation in the commit message.
  • Orphaned member references: When removing a member, their tasks in the Task Decomposition and CONFIG block must be reassigned or removed. Leaving orphaned assignees causes activation failures.
  • Wrong coordination pattern after evolution: Adding parallel-capable members to a sequential team, or making a hub-and-spoke team where agents need each other's output. Re-evaluate the pattern decision from create-team Step 4 after any structural change.
  • Team too large after adding members: Teams with more than 5 members become hard to coordinate. If evolution pushes the team past 5, consider splitting into two focused teams instead.
  • Stale See Also after variant creation: When creating a variant, both the original and the variant need to reference each other. One-directional references leave the graph incomplete.

Related Skills

  • create-team — foundation for authoring new teams; evolve-team assumes this was followed originally
  • evolve-skill — the parallel procedure for evolving SKILL.md files
  • evolve-agent — the parallel procedure for evolving agent definitions
  • commit-changes — commit the evolved team with a descriptive message

GitHub 仓库

pjt222/agent-almanac
路径: i18n/caveman-lite/skills/evolve-team
0
agentsagentskillsai-assisted-developmentclaude-codeskillsteams

相关推荐技能

content-collections

Content Collections 是一个 TypeScript 优先的构建工具,可将本地 Markdown/MDX 文件转换为类型安全的数据集合。它专为构建博客、文档站和内容密集型 Vite+React 应用而设计,提供基于 Zod 的自动模式验证。该工具涵盖从 Vite 插件配置、MDX 编译到生产环境部署的完整工作流。

查看技能

polymarket

这个Claude Skill为开发者提供完整的Polymarket预测市场开发支持,涵盖API调用、交易执行和市场数据分析。关键特性包括实时WebSocket数据流,可监控实时交易、订单和市场动态。开发者可用它构建预测市场应用、实施交易策略并集成实时市场预测功能。

查看技能

creating-opencode-plugins

该Skill帮助开发者创建OpenCode插件,用于接入命令、文件、LSP等25+种事件。它提供了插件结构、事件API规范和JavaScript/TypeScript实现模式,适合需要拦截操作、扩展功能或自定义事件处理的场景。开发者可通过它快速构建响应式模块来增强OpenCode AI助手的能力。

查看技能

sglang

SGLang是一个专为LLM设计的高性能推理框架,特别适用于需要结构化输出的场景。它通过RadixAttention前缀缓存技术,在处理JSON、正则表达式、工具调用等具有重复前缀的复杂工作流时,能实现极速生成。如果你正在构建智能体或多轮对话系统,并追求远超vLLM的推理性能,SGLang是理想选择。

查看技能