etanhey/skill-creator

Skill Creator

Take skills from "drafted" to "proven" through rigorous multi-layer evaluation.

Installation

bash ~/Gits/golems/skills/golem-powers/skill-creator/scripts/install.sh

The installer is idempotent — safe to re-run. It:

1. Symlinks the skill into ~/.claude/commands/skill-creator (golem-powers convention)
2. Ensures ~/Gits/skill-creator/.claude/agents/ and scripts/ exist as the project-scope home
3. Symlinks the session-miner agent definition into the project-scope agents dir (gates the sub-agent so only skillCreator-family sessions can spawn it)
4. Symlinks the parser into the project-scope scripts dir

Run bash scripts/install.sh --dry-run to preview without changes. bash scripts/install.sh --help for usage.

Core Loop: DRAFT → EVAL → RED → GREEN → SMOKE → SHIP

Every skill change goes through this pipeline. No exceptions.

1. DRAFT — Understand the Skill

• Read the SKILL.md or MCP tool definition completely
• brain_search for: prior evals, user complaints, known issues, related skills
• Identify: what does this skill claim to do? What's the baseline without it?
• Document intent, inputs, outputs, edge cases
• Classify: capability uplift (may become obsolete) vs encoded preference (durable)

2. EVAL DESIGN — Create Structured Test Cases

Design eval cases as structured JSON in evals/evals.json:

{
  "id": 1,
  "name": "descriptive-kebab-name",
  "category": "compliance|structure|quality",
  "description": "What this eval tests",
  "prompt": "The input scenario given to the agent",
  "assertions": [
    {
      "type": "tool_usage|content|negative",
      "name": "assertion-name",
      "description": "What correct behavior looks like"
    }
  ]
}

Cover: happy path, edge cases, failure modes, interaction with other skills. See references/scoring-rubric.md for scoring methodology.

3. RED — Baseline Measurement (without skill)

• Run each eval WITHOUT the skill loaded
• Score baseline performance across all eval cases
• Document baseline scores
• Gate: If baseline >70%, the skill may not be adding value — flag it

4. GREEN — With-Skill Measurement

• Run each eval WITH the skill loaded
• Score delta between with_skill and without_skill
• Gate: Delta <10% = marginal, consider retiring. Delta >30% = clearly valuable.
• Maximum 3 iterations before flagging for human review

5. SMOKE — Live Agent Testing

Two tiers based on skill importance:

Tier A: Static Smoke (all skills)

• Review evals.json assertions manually
• Verify no contradictions between skill instructions and assertions
• Check that strongest discriminators are tested

Tier B: Live Agent Smoke (flagship skills only)

• Run live-eval-runner.sh for top 3 discriminator evals
• Agent routing: Claude skills → Sonnet, code skills → Codex, audit skills → Cursor
• Compare captured output against assertions
• Gate: Live delta must be within 15% of static delta
• Results stored in evals/results/live-{date}.json and brain_store'd

See workflows/live-eval.md for the full live eval workflow.

6. SHIP — Package and Document

• Ensure every skill ships with its evals (no eval = no ship)
• brain_store eval results, delta scores, issues found
• Update skill metadata (version, last-eval-date, compliance-score)
• Follow /writing-skills for SKILL.md structure compliance
• REGISTER a NEW skill — committed ≠ installed. A new golem-powers skill is invisible to every agent until it's symlinked into ~/.claude/skills/<name> + ~/.claude/commands/<name> (run golem-install, which auto-discovers + links every golem-powers/*/ dir; or symlink the one new skill). Then VERIFY it appears in the available-skills list — committing/merging does NOT register it. (2026-05-30: /weave was committed + merged but unusable by any agent until the symlinks existed. See /writing-skills create.md "Final Step: REGISTER".)

Eval Methodology

with_skill vs without_skill comparison is MANDATORY. No exceptions.

Behavioral Compliance Scoring

| Weight | Dimension | What it measures | |--------|-----------|-----------------| | 70% | Compliance | Does the agent follow the skill's instructions? | | 20% | Structure | Does the output match expected format? | | 10% | Quality | Is the output actually good/useful? |

Scoring Thresholds

| Condition | Action | |-----------|--------| | Baseline >70% | Skill may not add value — flag and explain | | Delta <10% | Marginal — consider if complexity is worth it | | Delta >30% | Clearly valuable — ship with confidence | | Compliance <50% with skill | Instructions unclear — rewrite before shipping |

Agent Routing for Live Evals

| Skill type | Test with | Why | |------------|-----------|-----| | Claude behavior skills | Sonnet (default) | Tests actual Claude compliance | | Code implementation skills | Codex (default, no model flag) | Tests code quality | | Audit/review skills | Cursor (default) | Tests review thoroughness |

Workflows

| Workflow | When to use | |----------|-------------| | create-skill.md | Creating a new skill from scratch | | live-eval.md | Running live A/B tests with real agents | | ab-compare.md | Comparing skill versions or platforms | | mine-session.md | Mining a Claude Code session JSONL into a 10-section markdown digest (handoff docs, EOD waves, claim verification) |

References

| Reference | Content | |-----------|---------| | scoring-rubric.md | Full scoring methodology and rubric | | subagent-vs-skill.md | Classification reference — READ BEFORE SCAFFOLDING any new capability. Distinguishes skill (slash-triggered, parent-context) from sub-agent (name-spawned, isolated-context). Documents the misroute pattern (GitHub openai/codex#18823) that even Codex itself routinely makes. |

Integration with Other Skills

• /writing-skills — structural templates (SKILL.md format, frontmatter)
• /cmux-agents — spawning live eval agents in cmux panes
• /never-fabricate — Read() every file before reporting results
• /pr-loop — shipping skill changes through the full PR lifecycle

Packaged Sub-Agents (skill-creator family only)

These sub-agents ship in BOTH Claude and Codex formats. They are invokable only from sessions with cwd inside ~/Gits/skill-creator/ (i.e., skillCreatorClaude / skillCreatorCodex / skillCreatorRepoGolem). Sessions in other repos cannot spawn them directly — they must dispatch a skillCreator first.

Canonical source-of-truth lives in this skill's agents/ dir. scripts/install.sh symlinks them into the project repo's project-scope dirs (.claude/agents/ for Claude, .codex/agents/ for Codex). The dual-format packaging means the same sub-agent is available regardless of whether the parent is Claude Code or Codex CLI.

| Sub-agent | Claude format | Codex format | Workflow | |---|---|---|---| | session-miner | agents/session-miner.md (model: inherit) | agents/session-miner.toml (model: gpt-5.3-codex-spark) | mine-session.md |

Both formats invoke the same deterministic parser at scripts/session-miner.py. The Claude format uses subagent_type="session-miner" via the Agent tool; the Codex format is referenced by name in natural-language prompts (session_miner, mine X to Y).

For when to choose sub-agent vs skill: see references/subagent-vs-skill.md.

Critical Rules

1. Never claim "tests pass" without running them. Read actual output.
2. Every skill ships with evals. No eval = no ship.
3. brain_search BEFORE starting work — someone may have already evaluated it.
4. brain_store AFTER completing work — results, delta scores, decisions.
5. Max 3 iterations per skill before flagging for human review.
6. Read() every file you cite. No citing from compacted summaries.
7. Sequential skill compounding — build skills one at a time, each learning from the last.