skill-manager
Skill0 starsChanged 3 months agoApache-2.0View on GitHub
Coordinate the user's custom Claude Code skills for the current task — discover installed skills, understand what each does, and select the SMALLEST useful set relevant to the request while ignoring the rest. Detects conflicts, establishes a sensible execution order, honors explicit include/exclude instructions, and explains what was selected and why. Task-level selection is temporary — it does NOT modify, rename, or delete skill files unless the user explicitly asks for skill management. Use when the user asks which skills to use, wants skills coordinated for a task, or asks to create/edit/rename/delete a skill.
52 other projects use a version of this.
Adopted by
Most recently taken up by
- Adelie-Squad/solosquadtheir own version, first seen Nov 8, 2026
- jerry-guo-mys/skillstheir own version, first seen Nov 8, 2026
- Melvynx/ai-sprint-skillstheir own version, first seen Nov 8, 2026
- Melvynx/ai-maxxing-skillstheir own version, first seen Nov 1, 2026
- ZhangShixing0/skill-centertheir own version, first seen Nov 1, 2026
- thefrederiksen/devthrottletheir own version, first seen Nov 1, 2026
Used with
Projects with it also use Skill Evolution Manager, github-to-skills and 9 more.
Other files people use alongside this one.
Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
Arthur-devroleSkillSkill Evolution Manager
专门用于在对话结束时,根据用户反馈和对话内容总结优化并迭代现有 Skills 的核心工具。它通过吸取对话中的“精华”(如成功的解决方案、失败的教训、特定的代码规范)来持续演进 Skills 库。
Also in 5 projects that use this
Arthur-devroleSkillgithub-to-skills
Automated factory for converting GitHub repositories into specialized AI skills. Use this skill when the user provides a GitHub URL and wants to "package", "wrap", or "create a skill" from it. It automatically fetches repository details, latest commit hashes, and generates a standardized skill structure with enhanced metadata suitable for lifecycle management.
Also in 5 projects that use this
NVIDIASkillreview-code
Review a code repository for concrete issues in correctness, robustness, security, maintainability, tests, and integration. Use for repository-wide or focused code review where practical engineering judgment and strict scope control matter.
Also in 3 projects that use this
SentrySkillskill-creator
Create new agent skills following the Agent Skills specification. Use when asked to "create a skill", "add a new skill", "write a skill", "make a skill", "build a skill", or scaffold a new skill with SKILL.md. Guides through requirements, writing, registration, and verification.
Also in 12 projects that use this
OpenAISkillbrainstorming
You MUST use this before any creative work - creating features, building components, adding functionality, or modifying behavior. Explores user intent, requirements and design before implementation.
Also in 4 projects that use this
Microsoft AzureSkillhumanizer
Tone enforcement patterns for external-facing community responses
Also in 3 projects that use this
Other versions52 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 52
- anthropics/claude-for-legal★ 9.5ksame name
- buzhangsan/skill-manager★ 321same name
- thefrederiksen/cc-director★ 110same name
- thefrederiksen/devthrottle★ 110same name
- crisandrews/ClawCode★ 64same name
- Melvynx/ai-maxxing-skills★ 46same name
- egebese/skill-manager★ 38same name
- GeekFamilyCorp/TricorderKit★ 32same name
- Starchild-ai-agent/official-skills★ 32same name
- Adelie-Squad/solosquad★ 25same name
42 more not listed
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository49 other files it gives its agents.
- accessibility-auditorSkill
- api-doctorSkill
- browser-compatibilitySkill
- build-doctorSkill
- bundle-doctorSkill
- cache-doctorSkill
- change-cartographerSkill
- ci-cd-doctorSkill
- code-auditorSkill
- code-migratorSkill
- code-reviewerSkill
- config-doctorSkill
- container-doctorSkill
- context-keeperSkill
- contract-testerSkill
- cron-doctorSkill
- database-doctorSkill
- decision-memorySkill
- dependency-doctorSkill
- documentation-engineerSkill
- env-guardSkill
- error-boundary-doctorSkill
- feature-integritySkill
- git-hygieneSkill
- handoff-engineerSkill
- i18n-auditorSkill
- incident-debuggerSkill
- intent-architectSkill
- log-doctorSkill
- motion-designerSkill
- observability-engineerSkill
- performance-optimizerSkill
- product-architectSkill
- project-driftSkill
- qa-engineerSkill
- refactor-engineerSkill
- release-notesSkill
- runtime-doctorSkill
- safe-scopeSkill
- security-auditorSkill
- seo-doctorSkill
- ship-itSkill
- spec-doctorSkill
- state-managerSkill
- terminal-visualizerSkill
- test-engineerSkill
- type-safety-auditorSkill
- ui-redesignSkill
- ux-friction-hunterSkill
The file
We split this skill into the rules it gives an agent
12 rules. 0 are common, 11 are distinctive, and 1 is very specific to this project.
0
Common
Most of the 32 edited versions we read keep these.
11
Distinctive
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 44
Keep this in memory — do not write a file unless the user explicitly asks for a persistent registry.
- Distinctive: line 64
Look for conflicts, e.g.:
- Distinctive: line 79
When multiple skills are selected, give an explicit sequence tailored to the request.
- Distinctive: line 107
Treat as a task-level exclusion: do not apply that skill during this task.
- Distinctive: line 123
Only when the user explicitly asks to create, rename, edit, delete, improve, or change the description of a skill, help perform it.
- Very specific: line 135
If one skill suffices, use one.
- Distinctive: line 141
When skills overlap, resolve in this order:
- Distinctive: line 146
Use the skill most directly related to the task.
- Distinctive: line 147
Prefer specialized skills over generic ones.
- Distinctive: line 149
Use the minimum required skill set.
- Distinctive: line 155
When skill selection is requested, respond using exactly:
- Distinctive: line 172
Report honestly: do not claim a skill was *executed* if it was only *selected*.
“Common” is measured: kept by at least 6 in 10 of the 32 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Skill Manager
Coordinate the user's custom Claude Code skills for the task at hand. Discover what's installed, understand each skill's purpose, and select only the skills that materially help the current request — then explain the choice.
Core principle: use the minimum number of skills necessary to do the task correctly. Do not activate every available skill just because it exists. A skill is selected only when its purpose materially advances this task.
Hard boundary — selection is not modification. This skill reads skill files to coordinate them. It must not modify, delete, rename, or rewrite any skill file unless the user explicitly asks for skill management/editing (Phase 11). Selecting, excluding, or "disabling for this task" are temporary, task-level decisions that never touch the files on disk.
Also, always: never expose secrets; never modify unrelated projects.
Phase 1 — Discover skills
Inspect the available skills across all supported locations (read-only):
- Project-level:
./.claude/skills/*/SKILL.md(here:D:\claude\.claude\skills\). - User-level:
~/.claude/skills/*/SKILL.md(here:C:\Users\neera\.claude\skills\), including symlinked skill dirs. - Plugin/marketplace skills: under
~/.claude/plugins/.../skills/*/SKILL.md. - Any other supported skill location the environment exposes (the harness's available-skills list is also authoritative).
For each skill, read its frontmatter and enough of the body to understand: name · description · purpose · required tools · scope · potential conflicts, and which workflow stage it belongs to (planning, implementation, design, testing, auditing, deployment, documentation, or other). Do not modify anything.
Phase 2 — Build a skill registry (in memory)
For each discovered skill, build an internal entry:
Skill:
Purpose:
Category: (planning | design | implementation | testing | audit | deployment | docs | meta | other)
Best used when:
Avoid when:
Dependencies:
Potential conflicts:Keep this in memory — do not write a file unless the user explicitly asks for a persistent registry.
Phase 3 — Understand the current task
Analyze the request and determine: task type · project type · required workflow · complexity · and whether it involves planning, implementation, UI, testing, security, deployment, and whether documentation is required. This classification drives selection — don't select against a stage the task doesn't touch.
Phase 4 — Select skills
Choose the smallest useful set. For each selected skill:
✓ <skill-name>
Reason:
When it should run:For everything else: do not activate or apply it. Never force an unrelated skill into the workflow to seem thorough.
Phase 5 — Detect conflicts
Look for conflicts, e.g.:
- Two skills redesigning the same UI differently.
- Two skills proposing different architectures.
- A read-only skill conflicting with a modification request.
- A planning skill applied after implementation already started.
- A deployment skill used before testing.
- A cleanup skill running while another skill is modifying files.
Resolve conflicts by establishing a clear order. When appropriate, prefer:
Planning → Implementation → Testing → Audit → Deployment
Phase 6 — Create execution order
When multiple skills are selected, give an explicit sequence tailored to the request. Illustrative (do not apply blindly to every task):
/product-architect2./context-keeper3. Implementation 4./ui-redesign5./qa-engineer6./code-auditor7./ship-it
Order by the actual task — e.g. context-keeper first in an unfamiliar codebase; skip stages the task doesn't need.
Phase 7 — Recommendation mode ("Which skills should I use?")
Return a concise recommendation — selected and not-selected, each with a one-line reason:
Selected:
- /context-keeper
- /ui-redesign
- /qa-engineer
Not selected:
- /ship-it — no deployment requested
- /product-architect — architecture already exists
- /code-auditor — not needed for this taskPhase 8 — Explicit inclusion ("Use only X and Y")
Respect it exactly. Do not add other skills — unless a critical dependency makes the task impossible without one. If so, explain the dependency and get agreement before proceeding, rather than silently expanding the set.
Phase 9 — Deselection ("Don't use X")
Treat as a task-level exclusion: do not apply that skill during this task. Do not delete or disable the skill itself — it remains installed and available for future tasks.
Phase 10 — Temporary skill set
Support explicit task-level configs:
USE: DO NOT USE:
- ui-redesign - product-architect
- qa-engineer - ship-it / code-auditorThis is a temporary execution configuration for the current task only. Do not permanently modify the user's skill installation.
Phase 11 — Skill creation & management (only when explicitly asked)
Only when the user explicitly asks to create, rename, edit, delete, improve, or change the description of a skill, help perform it. Before modifying an existing skill:
- Read the existing skill fully.
- Understand its current purpose.
- Preserve useful behavior.
- Make the smallest necessary change.
- Verify the resulting frontmatter (valid
name/description/tools) and instructions.
Never silently overwrite a skill. For deletion, confirm the target and never remove a skill without that explicit instruction. Follow the repo's existing SKILL.md conventions when creating or editing.
Phase 12 — Avoid skill overload
If one skill suffices, use one. A simple UI color tweak should trigger only /ui-redesign — not product-architect, code-auditor, qa-engineer, and ship-it. A deployment-only task should not pull in UI skills. Match the set to the real scope.
Skill priority rules
When skills overlap, resolve in this order:
- Explicit user instruction — highest priority.
- Project-specific instructions (CLAUDE.md / project config).
- Safety constraints — always apply.
- Use the skill most directly related to the task.
- Prefer specialized skills over generic ones.
- Avoid duplicate workflows.
- Use the minimum required skill set.
Final output
When skill selection is requested, respond using exactly:
## Selected Skills
- `/skill-name` — reason
## Not Selected
- `/skill-name` — reason
## Execution Order
1. `/skill-name`
2. `/skill-name`
## Why
(brief: how the selected skills work together)Report honestly: do not claim a skill was executed if it was only selected; do not claim a skill was disabled if it was merely excluded from this task.
Guardrails, always: never delete skills without explicit permission; never modify skill files just to select/deselect them (selection is temporary and task-specific); never expose secrets; never modify unrelated projects; don't auto-activate every available skill; prefer the smallest effective set; and clearly distinguish "selected for this task" from "permanently disabled."
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

