Skill Creator

---
name: skill-creator
description: Create, improve, and evaluate skills using realistic test cases, baseline comparisons, and user feedback.
disable-model-invocation: true
---

# Skill Creator

> Adapted for Her from Anthropic Skill Creator (Apache-2.0); see UPSTREAM.md for the pinned source and changes.

A skill for creating new skills and iteratively improving them.

At a high level, the process of creating a skill goes like this:

- Decide what you want the skill to do and roughly how it should do it
- Write a draft of the skill
- Create a few test prompts and run Her with access to the skill on them
- Help the user evaluate the results both qualitatively and quantitatively
  - While the runs happen in the background, draft some quantitative evals if there aren't any (if there are some, you can either use as is or modify if you feel something needs to change about them). Then explain them to the user (or if they already existed, explain the ones that already exist)
  - Use the `eval-viewer/generate_review.py` script to show the user the results for them to look at, and also let them look at the quantitative metrics
- Rewrite the skill based on feedback from the user's evaluation of the results (and also if there are any glaring flaws that become apparent from the quantitative benchmarks)
- Repeat until you're satisfied
- Expand the test set and try again at larger scale

Your job when using this skill is to figure out where the user is in this process and then jump in and help them progress through these stages. So for instance, maybe they're like "I want to make a skill for X". You can help narrow down what they mean, write a draft, write the test cases, figure out how they want to evaluate, run all the prompts, and repeat.

On the other hand, maybe they already have a draft of the skill. In this case you can go straight to the eval/iterate part of the loop.

Of course, you should always be flexible and if the user is like "I don't need to run a bunch of evaluations, just vibe with me", you can do that instead.

Then after the skill is done (but again, the order is flexible), you can also run the skill description improver, using fresh Her sessions, to optimize the triggering of the skill.

Cool? Cool.

## Her environment

This is an explicitly invoked built-in skill: use `/skill:skill-creator` or the desktop's Create in a Conversation action. Its instructions and resources work in both desktop and terminal sessions.

- Create personal skills at `${HER_HOME:-~/.her}/skills/<name>/SKILL.md`. Only use `<project>/.agents/skills/<name>/SKILL.md` when the user wants a repository skill. Never edit this installed `.system` copy; copy an existing built-in into a user-owned directory before improving it.
- Use this session's existing model and tools. Do not invoke Claude Code, install another model SDK, request Anthropic credentials, or bypass Her's permissions to run evaluations.
- Python 3.10+ is optional for the bundled statistics, review, validation, and packaging scripts. They use only the standard library. Without Python, draft skills, run cases with the available tools, and review outputs in the conversation.
- For helper scripts, first check for the Her executable. `HER_BIN` can name the desktop's bundled executable or another known absolute path when `her` is not on PATH. Do not install the terminal CLI just to write a skill.
- Validate using Her's own parser: `python3 <skill-creator-path>/scripts/quick_validate.py <skill-folder>`. If the CLI is unavailable, review the header and references directly and check the installed skill in the desktop list after its watcher refreshes. Do not claim automated validation ran when it did not.
- New conversations discover new skills; existing conversations retain their initial skill set. Use `/skill:<name>` to explicitly invoke a discovered skill. Her supports `name`, `description`, `disable-model-invocation`, and `argument-hint`; `allowed-tools` does not grant or restrict tool access.

## Communicating with the user

The skill creator is liable to be used by people across a wide range of familiarity with coding jargon. If you haven't heard (and how could you, it's only very recently that it started), there's a trend now where the power of Her is inspiring plumbers to open up their terminals, parents and grandparents to google "how to install npm". On the other hand, the bulk of users are probably fairly computer-literate.

So please pay attention to context cues to understand how to phrase your communication! In the default case, just to give you some idea:

- "evaluation" and "benchmark" are borderline, but OK
- for "JSON" and "assertion" you want to see serious cues from the user that they know what those things are before using them without explaining them

It's OK to briefly explain terms if you're in doubt, and feel free to clarify terms with a short definition if you're unsure if the user will get it.

---

## Creating a skill

### Capture Intent

Start by understanding the user's intent. The current conversation might already contain a workflow the user wants to capture (e.g., they say "turn this into a skill"). If so, extract answers from the conversation history first — the tools used, the sequence of steps, corrections the user made, input/output formats observed. Ask only for gaps that affect the result; use answers already present in the conversation.

1. What should this skill enable Her to do?
2. When should this skill trigger? (what user phrases/contexts)
3. What's the expected output format?
4. Should we set up test cases to verify the skill works? Skills with objectively verifiable outputs (file transforms, data extraction, code generation, fixed workflow steps) benefit from test cases. Skills with subjective outputs (writing style, art) often don't need them. Suggest the appropriate default based on the skill type, but let the user decide.

### Interview and Research

Proactively ask questions about edge cases, input/output formats, example files, success criteria, and dependencies. Wait to write test prompts until you've got this part ironed out.

Check available MCPs - if useful for research (searching docs, finding similar skills, looking up best practices), research in parallel via subagents if available, otherwise inline. Come prepared with context to reduce burden on the user.

### Write the SKILL.md

Based on the user interview, fill in these components:

- **name**: Skill identifier
- **description**: When to trigger, what it does. This is the primary triggering mechanism - include both what the skill does AND specific contexts for when to use it. All "when to use" info goes here, not in the body. Keep the description precise enough to distinguish this skill from adjacent tasks; avoid expanding its scope just to increase trigger rates. - **compatibility**: Required tools, dependencies (optional, rarely needed)
- **the rest of the skill :)**

### Skill Writing Guide

#### Anatomy of a Skill

```
skill-name/
├── SKILL.md (required)
│   ├── YAML frontmatter (name, description required)
│   └── Markdown instructions
└── Bundled Resources (optional)
    ├── scripts/    - Executable code for deterministic/repetitive tasks
    ├── references/ - Docs loaded into context as needed
    └── assets/     - Files used in output (templates, icons, fonts)
```

#### Progressive Disclosure

Skills use a three-level loading system:
1. **Metadata** (name + description) - Always in context (~100 words)
2. **SKILL.md body** - In context whenever skill triggers (<500 lines ideal)
3. **Bundled resources** - As needed (unlimited, scripts can execute without loading)

These word counts are approximate and you can feel free to go longer if needed.

**Key patterns:**
- Keep SKILL.md under 500 lines; if you're approaching this limit, add an additional layer of hierarchy along with clear pointers about where the model using the skill should go next to follow up.
- Reference files clearly from SKILL.md with guidance on when to read them
- For large reference files (>300 lines), include a table of contents

**Domain organization**: When a skill supports multiple domains/frameworks, organize by variant:
```
cloud-deploy/
├── SKILL.md (workflow + selection)
└── references/
    ├── aws.md
    ├── gcp.md
    └── azure.md
```
Her reads only the relevant reference file.

#### Principle of Lack of Surprise

This goes without saying, but skills must not contain malware, exploit code, or any content that could compromise system security. A skill's contents should not surprise the user in their intent if described. Don't go along with requests to create misleading skills or skills designed to facilitate unauthorized access, data exfiltration, or other malicious activities. Things like a "roleplay as an XYZ" are OK though.

#### Writing Patterns

Prefer using the imperative form in instructions.

**Defining output formats** - You can do it like this:
```markdown
## Report structure
ALWAYS use this exact template:
# [Title]
## Executive summary
## Key findings
## Recommendations
```

**Examples pattern** - It's useful to include examples. You can format them like this (but if "Input" and "Output" are in the examples you might want to deviate a little):
```markdown
## Commit message format
**Example 1:**
Input: Added user authentication with JWT tokens
Output: feat(auth): implement JWT-based authentication
```

### Writing Style

Try to explain to the model why things are important in lieu of heavy-handed musty MUSTs. Use theory of mind and try to make the skill general and not super-narrow to specific examples. Start by writing a draft and then look at it with fresh eyes and improve it.

### Test Cases

After writing the skill draft, come up with 2-3 realistic test prompts — the kind of thing a real user would actually say. Share them with the user: [you don't have to use this exact language] "Here are a few test cases I'd like to try. Do these look right, or do you want to add more?" Then run them.

Save test cases to `evals/evals.json`. Don't write assertions yet — just the prompts. You'll draft assertions in the next step while the runs are in progress.

```json
{
  "skill_name": "example-skill",
  "evals": [
    {
      "id": 1,
      "prompt": "User's task prompt",
      "expected_output": "Description of expected result",
      "files": []
    }
  ]
}
```

See `references/schemas.md` for the full schema (including the `assertions` field, which you'll add later).

## Running and evaluating test cases

This section is one continuous sequence — don't stop partway through. Use the actual available Her tools; do not assume another testing skill is installed.

Put results in `<skill-name>-workspace/` as a sibling to the skill directory. Within the workspace, organize results by iteration (`iteration-1/`, `iteration-2/`, etc.) and within that, each test case gets a directory (`eval-0/`, `eval-1/`, etc.). Don't create all of this upfront — just create directories as you go.

### Step 1: Run independent with-skill and baseline cases

When `spawn_agent` is available, start separate agents for the with-skill and baseline cases. Her accepts `spawn_agent {description, task, name, wait}`; leave `wait` false so the cases run at the same time, and end your turn to collect their reports, which come back as messages. Respect available agent capacity. Subagents cannot spawn further agents, so the main session coordinates grading. If independent agents are unavailable, run sequentially and disclose the weaker comparison instead of presenting it as independent evidence.

**With-skill run:**

```
Execute this task:
- Skill path: <path-to-skill>
- Task: <eval prompt>
- Input files: <eval files if any, or "none">
- Save outputs to: <workspace>/iteration-<N>/eval-<ID>/with_skill/run-1/outputs/
- Save an evidence-based transcript to: <workspace>/iteration-<N>/eval-<ID>/with_skill/run-1/transcript.md
- Outputs to save: <what the user cares about — e.g., "the .docx file", "the final CSV">
```

**Baseline run** (same prompt, but the baseline depends on context):

Her agents share a filesystem. Keep output directories separate and ensure a baseline cannot discover or read the candidate skill; if the available tools cannot isolate discovery, state that limitation and do not label the result a clean no-skill baseline.
- **Creating a new skill**: no skill at all. Same prompt, no skill path, save to `without_skill/run-1/outputs/`.
- **Improving an existing skill**: the old version. Before editing, snapshot the skill (`cp -r <skill-path> <workspace>/skill-snapshot/`), then point the baseline subagent at the snapshot. Save to `old_skill/run-1/outputs/`.

Write an `eval_metadata.json` for each test case (assertions can be empty for now). Give each eval a descriptive name based on what it's testing — not just "eval-0". Use this name for the directory too. If this iteration uses new or modified eval prompts, create these files for each new eval directory — don't assume they carry over from previous iterations.

```json
{
  "eval_id": 0,
  "eval_name": "descriptive-name-here",
  "prompt": "The user's task prompt",
  "assertions": []
}
```

### Step 2: While runs are in progress, draft assertions

Don't just wait for the runs to finish — you can use this time productively. Draft quantitative assertions for each test case and explain them to the user. If assertions already exist in `evals/evals.json`, review them and explain what they check.

Good assertions are objectively verifiable and have descriptive names — they should read clearly in the benchmark viewer so someone glancing at the results immediately understands what each one checks. Subjective skills (writing style, design quality) are better evaluated qualitatively — don't force assertions onto things that need human judgment.

Update the `eval_metadata.json` files and `evals/evals.json` with the assertions once drafted. Also explain to the user what they'll see in the viewer — both the qualitative outputs and the quantitative benchmark.

### Step 3: As runs complete, capture timing data

When measured token usage and elapsed time are available, normalize them into `timing.json` in each run directory:

```json
{
  "total_tokens": 84852,
  "duration_ms": 23332,
  "total_duration_seconds": 23.3
}
```

Her reports `agent_status.elapsed_ms` and `agent_status.usage.per_model`. Sum input, output, cache_read, and cache_write counts across models only when available. Save known measurements; omit unavailable fields and report them as unavailable, never estimate tokens from output characters. A task report alone may not expose usage. Keep timing.json next to grading.json inside run-1/.

### Step 4: Grade, aggregate, and launch the viewer

Once all runs are done:

1. **Grade each run** — spawn a grader subagent (or grade inline) that reads `agents/grader.md` and evaluates each assertion against the outputs. Save results to `grading.json` in each run-1/ directory. The grading.json expectations array must use the fields `text`, `passed`, and `evidence` (not `name`/`met`/`details` or other variants) — the viewer depends on these exact field names. For assertions that can be checked programmatically, write and run a script rather than eyeballing it — scripts are faster, more reliable, and can be reused across iterations.

2. **Aggregate into benchmark** — run the aggregation script from the skill-creator directory:
   ```bash
   python3 -m scripts.aggregate_benchmark <workspace>/iteration-N --skill-name <name>
   ```
   This produces `benchmark.json` and `benchmark.md` with pass_rate, time, and tokens for each configuration, with mean ± stddev and the delta. If generating benchmark.json manually, see `references/schemas.md` for the exact schema the viewer expects.
Put each with_skill version before its baseline counterpart.

3. **Do an analyst pass** — read the benchmark data and surface patterns the aggregate stats might hide. See `agents/analyzer.md` (the "Analyzing Benchmark Results" section) for what to look for — things like assertions that always pass regardless of skill (non-discriminating), high-variance evals (possibly flaky), and time/token tradeoffs.

4. **Generate the offline reviewer** with both qualitative outputs and quantitative data:
   ```bash
   python3 <skill-creator-path>/eval-viewer/generate_review.py \
     <workspace>/iteration-N --skill-name "my-skill" \
     --benchmark <workspace>/iteration-N/benchmark.json \
     --static <workspace>/iteration-N/review.html
   ```
   For iteration 2+, also pass `--previous-workspace <workspace>/iteration-<N-1>`.
   Open or link this local HTML file using available desktop/browser tools. The viewer works offline; spreadsheets are downloadable rather than rendered by an external library. If Python or a browser is unavailable, present the cases, files, and comparison directly in the conversation.
5. Let the user review outputs before revising again. The static viewer downloads feedback.json; read the file the user actually exported, or collect feedback in the conversation. Do not assume a feedback file exists or search unrelated private folders.

### What the user sees in the viewer

The "Outputs" tab shows one test case at a time:
- **Prompt**: the task that was given
- **Output**: the files the skill produced, rendered inline where possible
- **Previous Output** (iteration 2+): collapsed section showing last iteration's output
- **Formal Grades** (if grading was run): collapsed section showing assertion pass/fail
- **Feedback**: a textbox that auto-saves as they type
- **Previous Feedback** (iteration 2+): their comments from last time, shown below the textbox

The "Benchmark" tab shows the stats summary: pass rates, timing, and token usage for each configuration, with per-eval breakdowns and analyst observations.

Navigation is via prev/next buttons or arrow keys. When done, they click "Submit All Reviews" which saves all feedback to `feedback.json`.

### Step 5: Read the feedback

When the user tells you they're done, read `feedback.json`:

```json
{
  "reviews": [
    {"run_id": "eval-0-with_skill-run-1", "feedback": "the chart is missing axis labels", "timestamp": "..."},
    {"run_id": "eval-1-with_skill-run-1", "feedback": "", "timestamp": "..."},
    {"run_id": "eval-2-with_skill-run-1", "feedback": "perfect, love this", "timestamp": "..."}
  ],
  "status": "complete"
}
```

Only treat empty feedback as acceptance when the user submitted the review as complete. Focus your improvements on the test cases where the user had specific complaints.

---

## Improving the skill

This is the heart of the loop. You've run the test cases, the user has reviewed the results, and now you need to make the skill better based on their feedback.

### How to think about improvements

1. **Generalize from the feedback.** The big picture thing that's happening here is that we're trying to create skills that can be used a million times (maybe literally, maybe even more who knows) across many different prompts. Here you and the user are iterating on only a few examples over and over again because it helps move faster. The user knows these examples in and out and it's quick for them to assess new outputs. But if the skill you and the user are codeveloping works only for those examples, it's useless. Rather than put in fiddly overfitty changes, or oppressively constrictive MUSTs, if there's some stubborn issue, you might try branching out and using different metaphors, or recommending different patterns of working. It's relatively cheap to try and maybe you'll land on something great.

2. **Keep the prompt lean.** Remove things that aren't pulling their weight. Make sure to read the transcripts, not just the final outputs — if it looks like the skill is making the model waste a bunch of time doing things that are unproductive, you can try getting rid of the parts of the skill that are making it do that and seeing what happens.

3. **Explain the why.** Try hard to explain the **why** behind everything you're asking the model to do. Today's LLMs are *smart*. They have good theory of mind and when given a good harness can go beyond rote instructions and really make things happen. Even if the feedback from the user is terse or frustrated, try to actually understand the task and why the user is writing what they wrote, and what they actually wrote, and then transmit this understanding into the instructions. If you find yourself writing ALWAYS or NEVER in all caps, or using super rigid structures, that's a yellow flag — if possible, reframe and explain the reasoning so that the model understands why the thing you're asking for is important. That's a more humane, powerful, and effective approach.

4. **Look for repeated work across test cases.** Read the transcripts from the test runs and notice if the subagents all independently wrote similar helper scripts or took the same multi-step approach to something. If all 3 test cases resulted in the subagent writing a `create_docx.py` or a `build_chart.py`, that's a strong signal the skill should bundle that script. Write it once, put it in `scripts/`, and tell the skill to use it. This saves every future invocation from reinventing the wheel.

This task is pretty important (we are trying to create billions a year in economic value here!) and your thinking time is not the blocker; take your time and really mull things over. I'd suggest writing a draft revision and then looking at it anew and making improvements. Really do your best to get into the head of the user and understand what they want and need.

### The iteration loop

After improving the skill:

1. Apply your improvements to the skill
2. Rerun all test cases into a new `iteration-<N+1>/` directory, including baseline runs. If you're creating a new skill, the baseline is always `without_skill` (no skill) — that stays the same across iterations. If you're improving an existing skill, use your judgment on what makes sense as the baseline: the original version the user came in with, or the previous iteration.
3. Launch the reviewer with `--previous-workspace` pointing at the previous iteration
4. Wait for the user to review and tell you they're done
5. Read the new feedback, improve again, repeat

Keep going until:
- The user says they're happy
- The feedback is all empty (everything looks good)
- You're not making meaningful progress

---

## Advanced: Blind comparison

For situations where you want a more rigorous comparison between two versions of a skill (e.g., the user asks "is the new version actually better?"), there's a blind comparison system. Read `agents/comparator.md` and `agents/analyzer.md` for the details. The basic idea is: give two outputs to an independent agent without telling it which is which, and let it judge quality. Then analyze why the winner won.

This is optional, requires subagents, and most users won't need it. The human review loop is usually sufficient.

---

## Description Optimization

The description field in SKILL.md frontmatter is the primary mechanism that determines whether Her invokes a skill. After creating or improving a skill, offer to optimize the description for better triggering accuracy.

### Step 1: Generate trigger eval queries

Create 20 eval queries — a mix of should-trigger and should-not-trigger. Save as JSON:

```json
[
  {"query": "the user prompt", "should_trigger": true},
  {"query": "another prompt", "should_trigger": false}
]
```

The queries must be realistic and something a Her desktop or terminal user would actually type. Not abstract requests, but requests that are concrete and specific and have a good amount of detail. For instance, file paths, personal context about the user's job or situation, column names and values, company names, URLs. A little bit of backstory. Some might be in lowercase or contain abbreviations or typos or casual speech. Use a mix of different lengths, and focus on edge cases rather than making them clear-cut (the user will get a chance to sign off on them).

Bad: `"Format this data"`, `"Extract text from PDF"`, `"Create a chart"`

Good: `"ok so my boss just sent me this xlsx file (its in my downloads, called something like 'Q4 sales final FINAL v2.xlsx') and she wants me to add a column that shows the profit margin as a percentage. The revenue is in column C and costs are in column D i think"`

For the **should-trigger** queries (8-10), think about coverage. You want different phrasings of the same intent — some formal, some casual. Include cases where the user doesn't explicitly name the skill or file type but clearly needs it. Throw in some uncommon use cases and cases where this skill competes with another but should win.

For the **should-not-trigger** queries (8-10), the most valuable ones are the near-misses — queries that share keywords or concepts with the skill but actually need something different. Think adjacent domains, ambiguous phrasing where a naive keyword match would trigger but shouldn't, and cases where the query touches on something the skill does but in a context where another tool is more appropriate.

The key thing to avoid: don't make should-not-trigger queries obviously irrelevant. "Write a fibonacci function" as a negative test for a PDF skill is too easy — it doesn't test anything. The negative cases should be genuinely tricky.

### Step 2: Review with user

Present the eval set to the user for review using the HTML template:

1. Read the template from `assets/eval_review.html`
2. Replace the placeholders:
   - `__EVAL_DATA_PLACEHOLDER__` → the JSON array of eval items (no quotes around it — it's a JS variable assignment)
   - `__SKILL_NAME_PLACEHOLDER__` → the skill's name
   - `__SKILL_DESCRIPTION_PLACEHOLDER__` → the skill's current description
3. Write to a temporary HTML file and open or link it using the available desktop/browser tools. Escape inserted JSON for an HTML script context (for example, replace `<` with `\u003c`) and HTML-escape text placeholders.
4. The user can edit queries, toggle should-trigger, add/remove entries, then click "Export Eval Set"
5. Read the actual exported eval_set.json path provided by the user, or keep edits in the conversation.

This step matters — bad eval queries lead to bad descriptions.

### Step 3: Evaluate descriptions with Her

Use new independent Her sessions or subagents with the candidate skill discoverable and automatic invocation enabled. Send each query without an explicit `/skill:` mention: mentioning the name forces invocation and does not measure automatic selection. Inspect actual tool transcripts for reads of SKILL.md; ask an evaluator to report evidence rather than guessing from the final answer. If the current session's tools do not provide fresh automatic-discovery sessions or transcripts, review the description qualitatively and explain that trigger accuracy was not measured.

When enough cases justify optimization, split the examples into train and held-out test sets and compare several descriptions using the same configured model. Propose revisions in the current session, then run fresh cases again. Do not modify a user's installed skill during each trial; use isolated test copies. Do not overwrite the actual skill until the selected description has been reviewed. Report test-set results without promising that one model's scores generalize to all providers.

The upstream `run_loop.py`, `run_eval.py`, and `improve_description.py` are intentionally not bundled: they call Claude Code and inspect its event format, which does not evaluate Her.

### How skill triggering works

Understanding the triggering mechanism helps design better eval queries. Automatically invocable skills appear in Her's `available_skills` list (skills with `disable-model-invocation: true` are explicit-only) with their name + description, and Her decides whether to consult a skill based on that description. The selected model may choose not to read a skill for tasks it can easily handle on its own — simple, one-step queries like "read this PDF" may not trigger a skill even if the description matches perfectly, because Her can handle them directly with basic tools. Use representative multi-step or specialized queries to measure actual behavior rather than assuming that a matching description guarantees invocation.

This means your eval queries should be substantive enough that Her would actually benefit from consulting a skill. Simple queries like "read file X" are poor test cases — they won't trigger skills regardless of description quality.

### Step 4: Apply the result

Take the selected description from the measured results and update the skill's SKILL.md frontmatter. Show the user before/after and report the scores.

---

### Validate, Package, and Present

Validate with Her before sharing. If Python and the Her executable are available, package a .skill archive (a zip file supported by Her):

```bash
python3 -m scripts.package_skill <path/to/skill-folder>
```

After packaging, link the resulting `.skill` file. Install with `her skills add <archive>` or the desktop upload action when requested.

---

## Limited environments

When subagents, Python, a CLI executable, or a browser are unavailable, use the parts the session supports. Draft and improve skills with the normal file tools, run sequential checks if appropriate, and collect user feedback in the conversation. Distinguish a manual review from a measured independent benchmark. Preserve the existing skill name when improving it. Package only when the user needs an archive; a folder containing SKILL.md is already installable in Her.

---

## Reference files

The agents/ directory contains instructions for specialized subagents. Read them when you need to spawn the relevant subagent.

- `agents/grader.md` — How to evaluate assertions against outputs
- `agents/comparator.md` — How to do blind A/B comparison between two outputs
- `agents/analyzer.md` — How to analyze why one version beat another

The references/ directory has additional documentation:
- `references/schemas.md` — JSON structures for evals.json, grading.json, etc.

---

Repeating one more time the core loop here for emphasis:

- Figure out what the skill is about
- Draft or edit the skill
- Run Her with access to the skill on test prompts
- With the user, evaluate the outputs:
  - Create benchmark.json and run `eval-viewer/generate_review.py` to help the user review them
  - Run quantitative evals
- Repeat until you and the user are satisfied
- Package the final skill and return it to the user.


Good luck!