agent-loop-ext
Crash-resilient external agent loop with state persistence and CI/CD integration
Enable interactive design mode
$ npx -y skills add jmagly/aiwg --skill devkit-create-skill --agent claude-codeHow it fires
How this skill gets triggered: by you, by Claude, or both.
/devkit-create-skillContext preview
The summary Claude sees to decide when to auto-load this skill.
Enable interactive design mode
namespace: aiwg name: devkit-create-skill platforms: [all] description: Enable interactive design mode triggers: - "help me build a custom workflow" - "help me turn this repeated process into an AIWG workflow" - "decide whether this is a skill agent addon or docs recipe" - "skill architect"
Create a new skill that can be triggered by natural language patterns.
Skills differ from commands:
Skills are ideal for:
Verify `$ARGUMENTS` contains:
**Note**: Skills cannot be added to extensions because skills need standalone functionality.
Check target exists:
ls ~/.local/share/ai-writing-guide/agentic/code/addons/<target>/ # or ls ~/.local/share/ai-writing-guide/agentic/code/frameworks/<target>/
Guide the user through skill design:
**Skill Purpose**: > What does this skill do? (e.g., "applies voice profiles to content", "validates code quality")
**Trigger Phrases**: > What natural language phrases should activate this skill? > Examples: > - "apply voice", "use voice", "write in <voice> voice" > - "validate code", "check quality", "review code" > - "format document", "clean up", "prettify"
**Input Requirements**: > What input does this skill need? > - Content to process > - Configuration parameters > - File paths
**Output Format**: > What does this skill produce? > - Transformed content > - Validation report > - Recommendations
**Reference Materials**: > Does this skill need supporting documentation? > - Style guides > - Validation rules > - Example outputs
Run the CLI scaffolding tool:
node ~/.local/share/ai-writing-guide/tools/scaffolding/add-skill.mjs \ <name> \ --to <target>
<target>/skills/<name>/ ├── SKILL.md # Main skill definition └── references/ # Supporting documentation
The generated SKILL.md needs customization.
**REQUIRED frontmatter fields:**
--- name: <skill-name> description: <what this skill does — REQUIRED, non-empty> version: 1.0.0 priority: medium --- # <Skill Name> ## Trigger Phrases Activate this skill when the user says: - "<phrase 1>" - "<phrase 2>" - "<phrase 3>" ## Input This skill expects: - <input requirement 1> - <input requirement 2> ## Execution Process 1. <Step 1> 2. <Step 2> 3. <Step 3> ## Output This skill produces: - <output 1> - <output 2> ## Examples ### Example 1: <scenario> **User**: "<example trigger>" **Result**: <what happens> ### Example 2: <scenario> **User**: "<example trigger>" **Result**: <what happens>
If the skill needs supporting documentation:
# Create reference files touch <target>/skills/<name>/references/style-guide.md touch <target>/skills/<name>/references/validation-rules.md
The CLI tool automatically updates the manifest. Verify:
{
"skills": ["existing-skill", "<new-skill>"]
}Skill Created: <name> ───────────────────── Location: <target>/skills/<name>/ Created: ✓ SKILL.md ✓ references/ Manifest updated: <target>/manifest.json Next Steps: 1. Edit SKILL.md to define trigger phrases 2. Add execution process details 3. Create reference materials (if needed) 4. Test with natural language triggers
# Create skill in addon /devkit-create-skill code-formatter --to aiwg-utils # Create skill with interactive guidance /devkit-create-skill voice-apply --to voice-framework --interactive # Create skill in framework /devkit-create-skill requirement-tracer --to sdlc-complete
1. **Clear trigger phrases**: Use natural, conversational language 2. **Multiple triggers**: Provide several ways to invoke the skill 3. **Specific execution**: Document exactly what happens when triggered 4. **Good examples**: Show realistic usage scenarios 5. **Reference materials**: Include supporting docs when needed
Reusable project context and specialist workflows for the AI tools you already use. Plan software, coordinate specialist reviews, prepare campaigns, investigate incidents, organize research, curate media, and maintain operational knowledge.
Repo: jmagly/aiwg
Crash-resilient external agent loop with state persistence and CI/CD integration
Detect requests for iterative autonomous agent loops and route to the appropriate loop executor
Automatically execute tests when code-generating agents modify source files, enforcing the execute-before-return pattern
Enable agent loops to learn from similar past tasks and share patterns across loops
Query and manage the executable feedback debug memory
Execute tests on generated code and iterate until passing