From dstoic
Creates, modifies, and orchestrates Claude Code tools (skills, agents, scripts). Use when user requests creating, updating, improving, or editing any Claude Code extension — skills, sub-agents, slash commands, script wrappers. Triggers include "create/make/new skill/command/agent/script", "tool for X", "slash command", "sub-agent", file paths with /skills/, /agents/. For plugin version bumps and release metadata, route directly to edit-plugin.
npx claudepluginhub digital-stoic-org/agent-skills --plugin dstoicThis skill uses the workspace's default tool permissions.
**CRITICAL**: Always triage first, then act.
Creates isolated Git worktrees for feature branches with prioritized directory selection, gitignore safety checks, auto project setup for Node/Python/Rust/Go, and baseline verification.
Executes implementation plans in current session by dispatching fresh subagents per independent task, with two-stage reviews: spec compliance then code quality.
Dispatches parallel agents to independently tackle 2+ tasks like separate test failures or subsystems without shared state or dependencies.
CRITICAL: Always triage first, then act.
✅ [TYPE] because: pollution cost (~X tokens × Yfreq), context mode (main|fork), key factorgraph TD
A[Tool Request] --> AA{Plugin<br/>version/release?}
AA -->|Yes| PLUGIN[✅ edit-plugin]
AA -->|No| B{Deterministic<br/>shell only?}
B -->|Yes| C{AI needed to<br/>decide when/how?}
C -->|No| BASH[✅ Bash Script<br/>0 tokens]
C -->|Yes| D{Token<br/>budget?}
D -->|<500| WRAP[✅ Skill + scripts/]
D -->|>500| SKILL_REF[✅ Skill + reference.md]
B -->|No| E{Pollution<br/>cost?}
E -->|High >2000 tokens<br/>OR deep exploration| AGENT[✅ Sub-Agent]
E -->|Medium 500-2000| F{Context<br/>mode?}
F -->|Needs isolation| FORK[✅ Skill context:fork]
F -->|Needs main context| SKILL_BIG[✅ Skill + reference.md]
E -->|Low <500| G{Frequency?}
G -->|1+ per session| SKILL[✅ Skill]
G -->|Rare <1/10| DIRECT[❌ No tool needed]
style PLUGIN fill:#B2DFDB,stroke:#00897B,stroke-width:2px,color:#000
style BASH fill:#90EE90,stroke:#000,stroke-width:2px,color:#000
style SKILL fill:#FFD700,stroke:#000,stroke-width:2px,color:#000
style WRAP fill:#FFA500,stroke:#000,stroke-width:2px,color:#000
style SKILL_REF fill:#FFA500,stroke:#000,stroke-width:2px,color:#000
style SKILL_BIG fill:#87CEEB,stroke:#000,stroke-width:2px,color:#000
style FORK fill:#DDA0DD,stroke:#000,stroke-width:2px,color:#000
style AGENT fill:#DDA0DD,stroke:#000,stroke-width:2px,color:#000
style DIRECT fill:#FFB6C1,stroke:#000,stroke-width:2px,color:#000
Default: All skills are dual-invocable (both /name and model auto-invoke). disable-model-invocation: true is opt-out for rare edge cases.
💡 Before creating: consider running /search-skill to discover existing solutions.
pick-model skill for model selection| Type | Guide | Location |
|---|---|---|
| Skill | references/skill-guide.md | skills/name/SKILL.md |
| Agent | references/agent-guide.md | .claude/agents/name.md or plugin agents/ |
| Plugin | edit-plugin skill | plugin.json + marketplace.json |
| Bash | Direct guidance (scripts/, chmod +x) | Project scripts dir |
Skill frontmatter (quick ref — see references/skill-guide.md for full spec):
name — lowercase-hyphensdescription — triggers + use cases (max 1024 chars)context — main (default) | forkmodel — opus | sonnet | haikuallowed-tools, argument-hint, hooks — optionalAgent frontmatter (quick ref — see references/agent-guide.md for full spec):
name, description — requiredtools / disallowedTools — allowlist or denylistmodel — opus | sonnet | haiku | inherit (default)permissionMode, maxTurns, skills, mcpServers, memory, hooks, background, isolation — optionalSTOP — answer YES/NO before proceeding:
| Question | Skill | Agent |
|---|---|---|
| Q1: Pollution cost acceptable? (<500 tokens × freq for skill) | Required | N/A (isolated) |
| Q2: SKILL.md <500 tokens (with reference.md for overflow)? | Required | N/A |
| Q3: Specific capability, not a workflow bundle? | Required | Required |
| Q3a: Requires multi-step AI reasoning with isolation? | N/A | Required (any YES → agent) |
Skill: ANY NO → STOP. Recommend alternative (fork, agent, direct request). Agent: ALL NO on Q3a → STOP. Recommend skill or bash script instead.
| Operation | Guidance |
|---|---|
| ✅ Read-only | Parallelize freely |
| ⚠️ Writes (independent files) | Sequential OR Plan Mode first |
| ❌ Destructive / >3 files | Plan Mode MANDATORY |
See references/frameworks.md for edge cases, conversion guide, and extended examples.