Skip to content

tzachbon/srulik-toolkit / plugins/srulik-toolkit/skills/skill-creator

skill-creator

Verified

Create or edit a skill, improve its instructions or triggers, evaluate it on real tasks, or benchmark a revision. Every skill creation or edit uses create-plan and implement-plan, including small changes; evaluation compares actual outputs and includes human review.

references/authoring.md

View on GitHub
# Authoring a skill

Adapted from Anthropic's skill-creator at
`ca45fc496a3436251f191354091d557da31c15b5`; modified for Srulik handoffs and
portable authoring. See [Apache-2.0 terms](../LICENSE.txt).

Use this reference as design evidence in `create-plan`, then as execution
instructions inside the approved `implement-plan` task. Every creation or edit
needs both handoffs; reading this reference does not authorize a write.

1. Inspect the current conversation, target repository guidance, nearby skills,
   and actual examples. Resolve facts from evidence before asking the user.
   Capture what the skill enables, its trigger, expected artifact, boundaries,
   and independently observable success. Preserve an existing skill's name.
2. Define a concise `name` and discriminating `description`: what it does and
   when it applies. Separate genuine trigger branches from synonyms. Preserve
   invocation policy unless a change is explicitly required. Keep automatic
   discovery enabled by default; sensitivity is handled by execution authority,
   not by silently hiding a skill.
3. Write a capability-based `SKILL.md` with an ordered workflow, completion
   evidence, and important failure/stop conditions. Put substantial branch- or
   host-specific instructions in linked references with explicit reading
   conditions. Resolve relative paths from their owning document, not cwd.
4. Bundle scripts only for repetitive deterministic operations, assets only
   when outputs need them, and examples only when they change interpretation.
   Reuse installed capabilities and native platform features. Explain fragile
   constraints; remove boilerplate, speculative modes, and duplicated rules.
5. Use 2–3 realistic test prompts unless the approved plan chooses another
   justified set. Cover the primary behavior and a meaningful boundary. Share
   the prompts and observable expectations with the fixture owner before runs.
   Subjective quality needs human judgment; do not disguise it as an objective
   pass/fail assertion.
6. Follow [evaluation](evaluation.md), inspect outputs and transcripts, and
   generalize feedback instead of overfitting a single example. If several
   runs independently reinvent the same useful helper, consider bundling it
   within the approved scope. A new uncovered requirement returns to planning.

Minimum package:

```text
skill-name/
  SKILL.md      # name/description frontmatter and instructions
  references/   # only substantial conditional instructions
  scripts/      # only needed deterministic helpers
  assets/       # only resources used in outputs
```

The additional directories are optional. A copied upstream resource retains its
license/notices and a prominent modification notice. Validate authored scripts
through their public interface and the repository's actual check; never claim
that syntax/frontmatter checks prove behavioral quality.
reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.Stars: 471kUpdated Claude CodeCodexVerifiedInstall
diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness.Stars: 322kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 322kUpdated Claude CodeCodexVerifiedInstall
executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is availableStars: 322kUpdated Claude CodeCodexVerifiedInstall