Skill Creator

Skill Creator

A skill for creating new skills and iteratively improving them.

Choose The Needed Work

Use the conversation and existing artifacts to establish intent. For a focused correction, edit and run relevant structural or behavioral checks, then deliver. For a new or substantially changed workflow, use realistic cases that exercise its decisions. Use the full paired benchmark and viewer workflow below when the user requests comparison or the change's uncertainty justifies it; it is not a prerequisite for every skill edit. Package only when distribution is requested.

Existing authorization covers the requested authoring and ordinary validation. Ask only for a material missing decision or additional side effects, not for permission to advance each phase. Keep evaluations inside the authorized resources; use fixtures or drafts for sends, publishing, deletion, and production changes unless real execution is explicitly authorized.

Before host-sensitive work, read exactly one compatibility guide from references/compatibility/: claudecode.md, codex.md, or other.md. Read more than one only when packaging or evaluating the same skill across multiple hosts.

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 them in an agent session that has access to the skill
• 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 when a failure or unresolved concern warrants another change
• Expand the test set when the change affects behavior not covered by current cases

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, which we have a whole separate script for, to optimize the triggering of the skill.

Cool? Cool.

Communicating with the user

The skill creator is liable to be used by people across a wide range of familiarity with coding jargon. There's a trend now where the power of modern agents 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. Extract the workflow, tools, corrections, inputs, outputs, and boundaries from the current conversation and existing artifacts. Proceed when they are sufficient; ask only about gaps that materially change the skill's behavior or scope.

1. What should this skill enable the agent to do?
2. When should this skill trigger? (what user phrases/contexts)
3. What's the expected output format?
4. Which checks would expose a meaningful mistake? Choose validation proportional to the behavioral change. A wording correction need not start a benchmark; changes to authorization or external writes benefit from boundary cases.

Interview and Research

Inspect relevant examples and dependencies first. Infer reversible defaults and start drafting when the task is clear. Ask about an edge case only when its answer changes the intended behavior; do independent work while that question is pending.

Check available MCPs - if useful for research (searching docs, finding similar skills, looking up best practices), research in parallel via spawn subagents, 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: State the capability and specific contexts where it helps. Keep routing precise, including a near-miss boundary when overlap is likely; do not broaden triggers just to increase activation. Stay under 1024 characters for Codex-style validators and move workflow details into the body.
• compatibility: Required tools, dependencies (optional, rarely needed)
• the rest of the skill :)

If the target host is Codex/OpenAI, create or update agents/openai.yaml when the skill should appear cleanly in product UI, provide a default prompt, declare MCP dependencies, or control implicit invocation. See references/compatibility/codex.md.

Skill Writing Guide

Anatomy of a Skill

skill-name/
├── SKILL.md (required)
│   ├── YAML frontmatter (name, description required)
│   └── Markdown instructions
├── agents/ (optional, host-specific)
│   └── openai.yaml - Codex/OpenAI UI metadata, default prompt, dependencies, and invocation policy
└── 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, but treat the metadata budget as real. Long descriptions can make routing worse or fail host validation. Keep the frontmatter description concise, and do not exceed 1024 characters.

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

The agent 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:

## 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):

## 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.

{
  "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).

Choose the skill location first

Before you create files, decide whether this skill is global or project-local.

• Global skills live in ~/.agents/skills/<skill-name>.
• Project-local skills live in <project-root>/.agents/skills/<skill-name>.
• Put the eval workspace next to the skill directory as <skill-name>-workspace/.
• If the user wants a global skill to be discoverable by Codex immediately, also create ~/.codex/skills/<skill-name> as a symlink to the global skill directory.

When running a benchmark, establish its workspace before writing benchmark files so subsequent paths are stable. This setup is not required for an ordinary skill edit.

Running and evaluating test cases

This section is one continuous sequence — don't stop partway through. Do NOT use /skill-test or any other testing skill.

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: Spawn all runs (with-skill AND baseline) in the same turn

For each paired test case, run one candidate and one baseline with equivalent inputs, model, and resources. Run pairs concurrently when agent capacity and isolation permit; otherwise use fresh isolated sequential runs and report the scheduling limit. Do not expand concurrency merely to satisfy a fixed count.

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/outputs/
- 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):

• Creating a new skill: no skill at all. Same prompt, no skill path, save to without_skill/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/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.

{
  "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 each subagent task completes, you receive a notification containing total_tokens and duration_ms. Save this data immediately to timing.json in the run directory:

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

This is the only opportunity to capture this data — it comes through the task notification and isn't persisted elsewhere. Process each notification as it arrives rather than trying to batch them.

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 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:

 python -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.

1. 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.
2. Launch the viewer with both qualitative outputs and quantitative data:

 nohup python <skill-creator-path>/eval-viewer/generate_review.py \
   <workspace>/iteration-N \
   --skill-name "my-skill" \
   --benchmark <workspace>/iteration-N/benchmark.json \
   > /dev/null 2>&1 &
 VIEWER_PID=$!

For iteration 2+, also pass --previous-workspace <workspace>/iteration-<N-1>. For a full benchmark, include a reviewable viewer artifact when supported. A focused validation can report its evidence directly. If the viewer is unavailable, return the inspectable outputs and limitation rather than withholding the completed evaluation. The viewer expects each run directory to contain an outputs/ subdirectory. If a lightweight or synthetic benchmark only produced grading.json and timing.json, create a minimal outputs/summary.md (or similar small artifact) before calling generate_review.py so the viewer has something to render. Headless environments: If webbrowser.open() is not available or the environment has no display, use --static <output_path> to write a standalone HTML file instead of starting a server. Feedback will be downloaded as a feedback.json file when the user clicks "Submit All Reviews". After download, copy feedback.json into the workspace directory for the next iteration to pick up.

Note: please use generate_review.py to create the viewer; there's no need to write custom HTML.

1. Tell the user something like: "I've opened the results in your browser. There are two tabs — 'Outputs' lets you click through each test case and leave feedback, 'Benchmark' shows the quantitative comparison. When you're done, come back here and let me know." If you used --static, include the exact HTML path or link in your response so the user can open it immediately. Do not make the user ask for the viewer after the fact.

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:

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

Empty feedback means the user thought it was fine. Focus your improvements on the test cases where the user had specific complaints.

Kill the viewer server when you're done with it:

kill $VIEWER_PID 2>/dev/null

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 affected cases and any shared behavior at risk. Keep the baseline explicit: no skill for a new skill, or an identified prior version for an existing one. Expand to the full suite when the change invalidates its earlier evidence.
3. For a full benchmark, launch the reviewer with --previous-workspace pointing at the previous iteration when supported.
4. Deliver the result after relevant checks pass. Wait for review only when the user requested a review checkpoint or a remaining decision needs their judgment.
5. Incorporate feedback when it arrives; do not require a satisfaction response to finish an otherwise completed edit.

Finish when:

• The requested scope is implemented and relevant validation supports it
• The user pauses the work
• A specific dependency prevents further useful progress; preserve and label the partial result

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 the Agent invokes a skill. After creating or improving a skill, offer to optimize the description for better triggering accuracy.

Do not optimize by bloating the description. Every candidate description should fit the same compatibility budget as a hand-written skill: usually 50-100 words, under 1024 characters, no angle brackets, and focused on task ownership, trigger phrases, near-miss boundaries, and the most important competing-skill exclusions.

Step 1: Generate trigger eval queries

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

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

The queries must be realistic and something a real user of the target Agent 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 temp file (e.g., /tmp/eval_review_<skill-name>.html) and open it: open /tmp/eval_review_<skill-name>.html
4. The user can edit queries, toggle should-trigger, add/remove entries, then click "Export Eval Set"
5. The file downloads to ~/Downloads/eval_set.json — check the Downloads folder for the most recent version in case there are multiple (e.g., eval_set (1).json)

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

Step 3: Run the optimization loop

Tell the user: "This will take some time — I'll run the optimization loop in the background and check on it periodically."

Save the eval set to the workspace, then run in the background:

python -m scripts.run_loop \
  --eval-set <path-to-trigger-eval.json> \
  --skill-path <path-to-skill> \
  --backend auto \
  --max-iterations 5 \
  --verbose

If the selected backend supports --model, use the model ID from your current session when practical so the tuning loop stays close to the user's experience.

While it runs, periodically tail the output to give the user updates on which iteration it's on and what the scores look like.

This handles the full optimization loop automatically. It splits the eval set into 60% train and 40% held-out test, evaluates the current description (running each query 3 times to get a reliable trigger rate), then calls the selected backend to propose improvements based on what failed. It re-evaluates each new description on both train and test, iterating up to 5 times. When it's done, it opens an HTML report in the browser showing the results per iteration and returns JSON with best_description — selected by test score rather than train score to avoid overfitting.

How skill triggering works

Understanding the triggering mechanism helps design better eval queries. On Agent's with native skill routing, skills appear in an available-skills list or equivalent routing context with their name + description, and the Agent's decides whether to consult a skill based on that description. The important thing to know is that Agent usually consult skills only for tasks they can't easily handle on their own — simple, one-step queries like "read this PDF" may not trigger a skill even if the description matches perfectly, because the base agent can handle them directly with basic tools. Complex, multi-step, or specialized queries reliably trigger skills when the description matches. On Codex, --backend codex is a judged proxy for this routing behavior, not a native trigger measurement.

This means your eval queries should be substantive enough that the Agent 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 best_description from the JSON output and update the skill's SKILL.md frontmatter. Show the user before/after and report the scores.

When applying the selected description to a target skill, keep it quoted in YAML and run a frontmatter parse check before reporting that the optimized skill is ready.

Package and Present (only if present_files tool is available)

Check whether you have access to the present_files tool. If you don't, skip this step. If you do, package the skill and present the .skill file to the user:

python -m scripts.package_skill <path/to/skill-folder>

After packaging, direct the user to the resulting .skill file path so they can install it.

Chat-only host instructions

In Claude.ai or another chat-only host, the core workflow is the same (draft → test → review → improve → repeat), but because these hosts don't have subagents, some mechanics change. Here's what to adapt:

Running test cases: No subagents means no parallel execution. For each test case, read the skill's SKILL.md, then follow its instructions to accomplish the test prompt yourself. Do them one at a time. This is less rigorous than independent subagents (you wrote the skill and you're also running it, so you have full context), but it's a useful sanity check — and the human review step compensates. Skip the baseline runs — just use the skill to complete the task as requested.

Reviewing results: If you can't open a browser (e.g., the host VM has no display, or you're on a remote server), skip the browser reviewer entirely. Instead, present results directly in the conversation. For each test case, show the prompt and the output. If the output is a file the user needs to see (like a .docx or .xlsx), save it to the filesystem and tell them where it is so they can download and inspect it. Ask for feedback inline: "How does this look? Anything you'd change?"

Benchmarking: Skip the quantitative benchmarking — it relies on baseline comparisons which aren't meaningful without subagents. Focus on qualitative feedback from the user.

The iteration loop: Improve the skill and rerun relevant cases. Deliver the evidence directly when a browser reviewer is unavailable; wait for feedback only if the user requested a checkpoint or a material decision remains. Iteration directories are useful for comparative runs, not required for every edit.

Description optimization: This section requires a shell-accessible backend such as claude or codex. Skip it if your current host doesn't expose one.

Blind comparison: Requires subagents. Skip it.

Packaging: The package_skill.py script works anywhere with Python and a filesystem. In a chat-only host, you can run it and the user can download the resulting .skill file.

Updating an existing skill: The user might be asking you to update an existing skill, not create a new one. In this case:

• Preserve the original name. Note the skill's directory name and name frontmatter field -- use them unchanged. E.g., if the installed skill is research-helper, output research-helper.skill (not research-helper-v2).
• Copy to a writeable location before editing. The installed skill path may be read-only. Copy to /tmp/skill-name/, edit there, and package from the copy.
• If packaging manually, stage in /tmp/ first, then copy to the output directory -- direct writes may fail due to permissions.

Headless worker host instructions

If you're in a headless worker host, the main things to know are:

• You have subagents, so the main workflow (spawn test cases in parallel, run baselines, grade, etc.) all works. (However, if you run into severe problems with timeouts, it's OK to run the test prompts in series rather than parallel.)
• You don't have a browser or display, so when generating the eval viewer, use --static <output_path> to write a standalone HTML file instead of starting a server. Then proffer a link that the user can click to open the HTML in their browser.
• For a full benchmark, generate the eval viewer with generate_review.py when available so the user can inspect examples. Focused checks can be reported directly. Review the evidence and fix clear errors without waiting for an additional user response.
• If your benchmark is lightweight and a run directory does not yet have outputs/, create a minimal artifact such as outputs/summary.md first. generate_review.py only discovers runs that have an outputs/ directory.
• Feedback works differently: since there's no running server, the viewer's "Submit All Reviews" button will download feedback.json as a file. You can then read it from there (you may have to request access first).
• Packaging works — package_skill.py just needs Python and a filesystem.
• Description optimization (run_loop.py / run_eval.py) should work in a headless worker host as long as the selected backend CLI is available, but please save it until you've fully finished making the skill and the user agrees it's in good shape.
• Updating an existing skill: The user might be asking you to update an existing skill, not create a new one. Follow the update guidance in the chat-only host section above.

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/compatibility/claudecode.md — Claude Code behavior, real trigger measurement, and baseline mechanics
• references/compatibility/codex.md — Codex/OpenAI behavior, judged trigger proxy, and agents/openai.yaml
• references/compatibility/other.md — Generic shell agents, chat-only hosts, and headless worker adaptations
• references/schemas.md — JSON structures for evals.json, grading.json, etc.

Deliver the changed skill and the checks actually performed. Distinguish structural validation, simulated boundary review, and real execution benchmarks; none implies the others. Include a viewer or package when the selected workflow calls for it.