creating-skills

# Skill Creator A skill for creating new skills and iteratively improving them. **Core principle: Creating skills IS Test-Driven Development applied to process documentation.** Follow RED-GREEN-REFACTOR: run baseline tests without the skill (RED - watch agent fail), write the skill addressing those failures (GREEN - watch agent comply), then close loopholes (REFACTOR - stay compliant). If you didn't watch an agent fail without the skill, you don't know if the skill teaches the right thing. 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 the agent 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, which we have a whole separate script for, to optimize the triggering of the skill. Cool? Cool. ## What Is a Skill A **skill** is a reference guide for proven techniques, patterns, or tools. Skills help future agent instances find and apply effective approaches. **Skills are:** Reusable techniques, patterns, tools, reference guides **Skills are NOT:** Narratives about how you solved a problem once ### Skill Types - **Technique**: Concrete method with steps to follow (e.g., condition-based-waiting, root-cause-tracing) - **Pattern**: Way of thinking about problems (e.g., flatten-with-flags, test-invariants) - **Reference**: API docs, syntax guides, tool documentation (e.g., office docs) ### When to Create a Skill **Create when:** - Technique wasn't intuitively obvious to you - You'd reference this again across projects - Pattern applies broadly (not project-specific) - Others would benefit **Don't create for:** - One-off solutions - Standard practices well-documented elsewhere - Project-specific conventions (put in AGENTS.md) - Mechanical constraints enforceable with regex/validation ## 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 the agent 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. The user may need to fill the gaps, and should confirm before proceeding to the next step. 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. 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. Use only letters, numbers, and hyphens. Prefer gerund form (verb + -ing) for processes: `creating-skills`, `testing-skills`. Name by what you DO or core insight, not generic terms. - **description**: When to trigger. This is the primary triggering mechanism. **CRITICAL: Description = When to Use, NOT What the Skill Does.** Start with "Use when..." and include specific symptoms, situations, and contexts. Write in third person. **NEVER summarize the skill's process or workflow** — when a description summarizes workflow, the agent may follow the description instead of reading the full skill content. Keep under 500 characters. All "when to use" info goes here, not in the body. Note: currently the agent tends to "undertrigger" skills -- to not use them when they'd be useful. To combat this, please make the skill descriptions a little bit "pushy". - Include concrete triggers: error messages, symptoms, tool names, file types - Use synonyms: "timeout/hang/freeze", "cleanup/teardown/afterEach" - Describe the *problem* not *language-specific symptoms* - Make sure to use this skill whenever the user mentions relevant keywords, even if they don't explicitly ask for the skill by name - **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 agent 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. **Incremental file writing (CRITICAL):** The model cannot output large files in a single call — each edit is limited to ~1000 tokens of output. Exceeding this causes API error 500. Always instruct agents to write large files incrementally: 1. Write the skeleton first (imports, signatures, section comments) 2. Fill in one section per edit call (~1000 tokens per chunk) 3. Final review and commit Include this pattern in any skill that generates large outputs (documents, code files, configs). Example instruction to include: ```markdown ## Writing Large Files This document may be large. Write it incrementally to avoid API errors: 1. **First pass:** Write the header, structure, and section outlines 2. **Subsequent passes:** Edit to fill in each section one at a time (~1000 tokens per edit) 3. **Final pass:** Review the complete document, fix issues, save ``` **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. ### Token Efficiency (Critical) The context window is a shared resource. Every token in SKILL.md competes with conversation history and other context. **Target word counts:** - Frequently-loaded skills: <200 words total - Other skills: <500 words (still be concise) **Techniques:** - **Default assumption: the agent is already very smart.** Only add context the agent doesn't already have. Challenge each piece: "Does the agent really need this explanation?" - **Move details to tool help:** Instead of documenting all flags, reference `--help` - **Use cross-references:** Reference other skills by name rather than repeating their workflows. Use `**REQUIRED SUB-SKILL:** Use skill-name` or `**REQUIRED BACKGROUND:** You MUST understand skill-name`. Never use `@` syntax — it force-loads files immediately, consuming context. - **Compress examples:** One excellent example beats many mediocre ones. Don't implement in 5+ languages. Don't create fill-in-blank templates. - **Eliminate redundancy:** Don't repeat what's in cross-referenced skills. Don't explain what's obvious from commands. Don't include multiple examples of the same pattern. - **Avoid time-sensitive information:** Don't include dates that will become outdated. Use "old patterns" sections for deprecated approaches. - **Use consistent terminology:** Choose one term and use it throughout. ### Code Examples **One excellent example beats many mediocre ones.** Choose the most relevant language for the task (testing techniques → TypeScript/JS, system debugging → Shell/Python, data processing → Python). A good example is complete, runnable, well-commented explaining WHY, from a real scenario, and ready to adapt. ### Flowchart Usage Use flowcharts (graphviz DOT format) ONLY for non-obvious decision points, process loops where the agent might stop too early, or "When to use A vs B" decisions. Never use flowcharts for reference material, code examples, linear instructions, or labels without semantic meaning. See `references/graphviz-conventions.dot` for graphviz style rules. Use `render-graphs.js` to render flowcharts to SVG for your human partner: ```bash ./render-graphs.js ../some-skill # Each diagram separately ./render-graphs.js ../some-skill --combine # All diagrams in one SVG ``` ### Anti-Patterns - **Narrative examples** ("In session 2025-10-03, we found...") — Too specific, not reusable - **Multi-language dilution** (example-js.js, example-py.py, example-go.go) — Mediocre quality, maintenance burden - **Code in flowcharts** — Can't copy-paste, hard to read - **Generic labels** (helper1, step3, pattern4) — Labels should have semantic meaning - **Too many options** — Don't present multiple approaches unless necessary. Provide a default with an escape hatch. - **Deeply nested references** — Keep file references one level deep from SKILL.md. The agent may partially read files referenced from other referenced files. - **Windows-style paths** — Always use forward slashes ### The Iron Law ``` NO SKILL WITHOUT A FAILING TEST FIRST ``` This applies to NEW skills AND EDITS to existing skills. Run a baseline scenario WITHOUT the skill before writing it. Watch the agent fail. Document exact rationalizations. Then write the skill addressing those specific failures. **No exceptions:** - Not for "simple additions" - Not for "just adding a section" - Not for "documentation updates" ### 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). ### Testing All Skill Types Different skill types need different test approaches: **Discipline-enforcing skills** (rules/requirements like TDD, verification-before-completion): - Test with pressure scenarios combining 3+ pressures: time, sunk cost, authority, exhaustion - Force explicit A/B/C choices, not open-ended questions - Identify rationalizations and add explicit counters - Success: Agent follows rule under maximum pressure **Technique skills** (how-to guides): - Test with application scenarios and edge cases - Test for missing information and gaps in instructions - Success: Agent successfully applies technique to new scenario **Pattern skills** (mental models): - Test recognition: Does agent know when pattern applies? - Test counter-examples: Does agent know when NOT to apply? - Success: Agent correctly identifies when/how to apply pattern **Reference skills** (documentation/APIs): - Test retrieval: Can agent find the right information? - Test application: Can agent use what they found correctly? - Gap testing: Are common use cases covered? - Success: Agent finds and correctly applies reference information ### Bulletproofing Against Rationalization For skills that enforce discipline, agents will find loopholes under pressure. Close every loophole explicitly: 1. **Explicit negation:** Don't just state the rule — forbid specific workarounds 2. **Rationalization table:** Capture every excuse agents make during testing 3. **Red flags list:** Make it easy for agents to self-check when rationalizing 4. **Foundational principle:** Add "Violating the letter of the rules is violating the spirit of the rules" to cut off "spirit vs letter" arguments See `references/persuasion-principles.md` for research on how authority, commitment, and scarcity principles increase compliance. ## Running and evaluating test cases **Before running any Python scripts, activate the opencode virtual environment:** ```bash source ~/.local/opencode-venv/bin/activate ``` This section is one continuous sequence — don't stop partway through. 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 test case, spawn two subagents in the same turn — one with the skill, one without. This is important: don't spawn the with-skill runs first and then come back for baselines later. Launch everything at once so it all finishes around the same time. **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 sub-agent 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. ```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 each sub-agent task completes, you receive a notification containing `total_tokens` and `duration_ms`. Save this data immediately to `timing.json` in the run directory: ```json { "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 the `grader` global agent using the Task tool (or grade inline) that 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: ```bash 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. 3. **Do an analyst pass** — use the `analyzer` global agent to read the benchmark data and surface patterns the aggregate stats might hide. Look for things like assertions that always pass regardless of skill (non-discriminating), high-variance evals (possibly flaky), and time/token tradeoffs.