a554b554/Reactant / skills/skill-generator
skill-generator
VerifiedReactant direct-entry skill authoring utility. Use when the user directly asks to create a project-local Reactant tag skill from a natural-language description; it has no inline tag form and is never dispatched by execute or execute-one.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 3.4 KB |
| references/skill-contract.md | markdown | 3.4 KB |
references/skill-contract.md
View on GitHub# Reactant Tag-Skill Contract (local excerpt)
A Reactant tag skill is a standard Agent Skills directory that the Reactant
router can dispatch for an inline tag. This excerpt is self-contained; the
full developer documentation lives in the Reactant package repository.
## Directory
```
<name>/
├── SKILL.md # required
├── scripts/ # optional executable helpers
├── references/ # optional extra documentation loaded on demand
└── assets/ # optional static resources
```
- `<name>`: lowercase letters/numbers separated by single hyphens. Must match
the `name` in SKILL.md frontmatter. Must not be one of the reserved names:
`execute`, `execute-one`, `skill-generator`, `chat-user`, `resolve`,
`chat-model`, `comment`.
- The directory must be self-contained: never reference files outside it.
## Activation boundary
The package entry skills are `execute`, `execute-one`, and `skill-generator`;
the generator itself has no inline tag form. Every generated skill is tag-only
and cannot become another entry skill. Its frontmatter description and an
`Activation Boundary` section must say that the worker runs only when
`execute` or `execute-one` selects its exact tag from a document. It must not
activate from ordinary chat-window requests, semantic similarity, or another
non-router skill. Without router handoff, it changes nothing.
## SKILL.md
YAML frontmatter with exactly these required fields, each a single line:
```
---
name: <name>
description: What the skill does and when to use it (at most 1024 characters).
---
```
The body must include, in host-neutral language:
1. **Activation Boundary** — the tag-only restriction above, including clean
refusal when no router handoff is present.
2. **Tag** — the inline form, shown literally: `<@<name>>` or
`<@<name>: prompt>`.
3. **Input** — what the router supplies: file path, surrounding block, prompt,
plus chain/context when relevant.
4. **Workflow** — numbered steps the agent follows.
5. **Scope modifiers** — how the skill treats `<<protect>>` (never modify the
content) and `((field))` (modify only inside), and which delimiters it
consumes on success.
6. **Context references** — double-backtick references in the prompt resolve
relative to the document's directory unless absolute; say whether each is
input or output.
7. **Consumption rule** — exactly what the skill replaces or removes on
success (usually its own tag; edit-like skills replace the surrounding
block). Everything else — unrelated tags, `<@comment: ...>`,
`<@chat-model>` blocks, and contextual literal examples — is preserved
byte-for-byte.
8. **Failure behavior** — on error or a missing host capability, leave the
tag in place, change nothing, and report. No retries, no second model
pass, no output verification: the result is accepted as produced.
## Validation
Run structural validation before considering the skill done. Check the Agent
Skills metadata, directory/name match, required Reactant contract sections,
self-contained resource paths, consumption rule, and clean failure behavior.
Developers may additionally use the optional `reactant lint-skill` command or
the official Agent Skills validator. Validation concerns skill/package
structure only; it must never scan an authored document or decide which text
is an actionable annotation. Fix every error and do not leave an invalid skill
in `.reactant/skills/`.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| 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: 469k | Updated | Claude CodeCodex | Verified | Install |
| 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: 321k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 321k | Updated | Claude CodeCodex | Verified | Install |
| 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 available | Stars: 321k | Updated | Claude CodeCodex | Verified | Install |