Skill Executor — Data & AI agent for Claude Code
Runs one task in a fresh context, with or without a skill, and saves everything it produced, for evals where the AI tool can't be run headless.
How to install Skill Executor
Installs to ~/.claude/agents/thenavidm-skill-builder-skill-executor.md
mkdir -p ~/.claude/agents && curl -fsSL https://raw.githubusercontent.com/thenavidm/skill-builder/HEAD/agents/skill-executor.md -o ~/.claude/agents/thenavidm-skill-builder-skill-executor.md Restart Claude Code, or start a new session, for it to be picked up.
What Skill Executor does
name: skill-executor description: Runs one task in a fresh context, with or without a skill, and saves everything it produced, for evals where the AI tool can't be run headless. Started by skill-eval, skill-bench and skill-compete.
You run one test task as a user would, with no memory of how the skill was made. You get the task's prompt, its input files, a folder to work in, and either a skill's path or the word "without".
- With a skill: read its
SKILL.mdand follow it to do the tas
Alternatives in Data & AI
- Claude Mem — A Claude Code plugin that automatically captures everything Claude does during your coding sessions, compresse 38.9k ★
- Credential Hunter — Autonomous credential-attack pipeline runner 4.5k ★
- Harness Scout — Picks the right AI agent harness for a described task or project 801 ★
Full documentation available on GitHub
View Source RepositoryRelated Agents
Slop Detector
Adversarial generic-design hunt with fresh context — finds structural, component, visual, copy, interaction, a
Eval Best No Type Rules
Eval arm best-no-type-rules (best without render_chart's per-type rules; Q23 arm C). Answers one building-data
Mp Goal Assessor
Read-only, fresh-context assessment of a completed masterplan run's declared goals. Consumes goals.md as QUOTE
Newsletter Competitive Analyst
Use this agent when you need to analyze competing AI newsletters to identify content gaps, differentiation opp
Architect Reviewer Complex
[COMPLEX tier (risk tier L) — opus/medium] Adversarial reviewer for a system design (design.md) or an ADR, wit
Scaffold Inspect
Sets up inspect-ai evaluation configurations for all runs in a designed experiment. Reads experiment_summary.y
Related Skills
Close Session Report
Run the full /close-session workflow, then write everything it produced (files changed, memory saved, TODO ite
Autopilot Nt
The executor — work a fix-workplan or prose goal to completion unattended in an isolated worktree; verify fres
Headless Harness Bench
Benchmark of 6 coding-agent harnesses (omp/pi/fx/opencode/dsh/crush) as headless agent loops driven by a contr