Autonomous task execution with library discovery, 3-tier validation, and atomic git commits. Integrates with Shannon CLI Python modules for platform-specific execution. Invokes /shannon:wave for code generation, validates outputs functionally, commits only validated changes. Use when: user requests autonomous execution, wants library-first development, needs validated commits.
Autonomous task execution with library discovery, 3-tier validation, and atomic git commits. Uses Shannon CLI to find existing libraries, validates changes functionally (build → tests → runtime), and commits only working code. Triggers on requests for "autonomous execution" or when user wants validated, library-first development with automatic git commits.
/plugin marketplace add krzemienski/shannon-framework/plugin install shannon@shannon-frameworkThis skill is limited to using the following tools:
references/FUNCTIONAL_VALIDATION_PROTOCOL.mdreferences/GIT_WORKFLOW_PROTOCOL.mdreferences/LIBRARY_DISCOVERY_PROTOCOL.mdreferences/README.mdExecute tasks autonomously with automatic library discovery, 3-tier functional validation, and atomic git commits. Provides a complete autonomous execution workflow that:
Key Innovation: Combines Shannon Framework's multi-agent orchestration (/shannon:wave) with Shannon CLI's validation and git automation for truly autonomous, validated execution.
MANDATORY when user says:
RECOMMENDED when:
DO NOT USE when:
Action: Invoke /shannon:prime for session setup
@skill context-preservation
operation: prepare
task_focused: true
Purpose:
Output: Session ready, skills discovered, context loaded
Action: Call Shannon CLI library discoverer
# Execute CLI command to search registries
shannon discover-libs "[feature extracted from task]" --category [ui|auth|networking|data|forms] --json
Purpose:
Output: List of LibraryRecommendation objects with install commands and reasoning
Example:
{
"libraries": [
{
"name": "next-auth",
"score": 95.0,
"why_recommended": "15k+ stars, maintained 4 days ago, MIT license",
"install_command": "npm install next-auth"
}
]
}
Integration with Phase 4: Library recommendations inject into wave execution context
Action: Invoke /shannon:analyze for complexity assessment
@skill spec-analysis
specification: {task description with context}
include_mcps: true
Purpose:
Output: Complexity score, domain breakdown, MCP recommendations
Usage Decision:
Action: Build execution plan with library context
For Simple Tasks (complexity <0.30):
// Single-step execution
const plan = {
steps: [{
number: 1,
description: task,
libraries: discoveredLibraries.map(l => l.name),
validation: {tier1: ["build"], tier2: ["test"], tier3: ["functional"]}
}]
}
For Complex Tasks (complexity ≥0.30):
// Use sequential-thinking MCP for multi-step planning
@mcp sequential-thinking
task: Break down {task} into execution steps
context: Recommended libraries: {libraries}
format: [{step: "", files: [], validation: {}}]
Output: ExecutionPlan with steps, each having validation criteria
FOR EACH step in plan:
Action: Invoke wave with enhanced prompts
@skill wave-orchestration
task: {step.description}
enhanced_context: |
RECOMMENDED LIBRARIES: {step.libraries}
CRITICAL INSTRUCTIONS:
- Use recommended libraries (don't build custom)
- Make minimal focused changes
- Follow project conventions
- Include error handling
[Full enhanced prompts from Shannon CLI]
Purpose:
Output: Files modified/created
Action: Extract file changes from wave execution
const filesChanged = parseWaveMessages(messages).filter(m =>
m.type === 'ToolUseBlock' && ['Write', 'Edit'].includes(m.name)
).map(m => m.input.file_path)
Output: List of files that were modified
Action: Call Shannon CLI validator
# Tier 1: Static validation
shannon validate --tier 1 --json
# If Tier 1 passes, Tier 2: Tests
shannon validate --tier 2 --json
# If Tier 2 passes, Tier 3: Functional
shannon validate --tier 3 --json
Validation Tiers:
Tier 1 - Static (~10s):
npm run build or cargo build or xcodebuildtsc --noEmit (if TypeScript) or mypy . (if Python)eslint or ruff check or swiftlintTier 2 - Tests (~1-5min):
npm test, pytest tests/, xcodebuild testTier 3 - Functional (~2-10min):
npm run dev, wait for server, curl health endpoint, verify 200 OKuvicorn main:app, curl endpoint, verify response correctOutput: ValidationResult(tier1_passed, tier2_passed, tier3_passed, all_passed)
IF validation.all_passed:
# Commit validated changes
shannon git-commit \
--step "{step.description}" \
--files "{filesChanged.join(',')}" \
--validation-json "{validation.toJSON()}"
Creates commit:
feat: {step description}
VALIDATION:
- Build: PASS
- Tests: 12/12 PASS
- Functional: Feature works in browser
Files: package.json, src/auth/login.tsx, ...
Result: Proceed to next step
ELSE validation failed:
# Rollback changes
shannon git-rollback
# Research solution (if research enabled)
@mcp firecrawl
search: "{validation.failures[0]} solution"
# Replan step with research findings
# Retry (attempt 2 of 3)
Result: Rollback → Research → Retry (max 3 attempts)
IF all attempts fail:
Action: Generate comprehensive report
const report = {
success: true,
task: taskDescription,
steps_completed: completedSteps.length,
steps_total: plan.steps.length,
commits_created: commits.map(c => c.hash),
branch_name: branchName,
duration_seconds: totalDuration,
libraries_used: uniqueLibraries,
validations: {
passed: validationsPassed,
failed: validationsFailed
}
}
Display:
✅ AUTONOMOUS EXECUTION COMPLETE
Task: {taskDescription}
Branch: {branchName}
Commits: {commits.length}
Duration: {duration}s
Libraries: {libraries.join(', ')}
Ready for: git push origin {branchName}
Save to Serena:
@mcp serena
write_memory: exec_result_{timestamp}
content: {report.toJSON()}
Technical:
Functional:
User Experience:
Symptom: Task builds custom authentication instead of using next-auth Cause: Phase 2 not executed or results not used Fix: Always run library discovery, inject results into wave context
Symptom: Commit created but code doesn't build Cause: Skipped validation or ignored failures Fix: Enforce all_passed check before commit
Symptom: Git history has broken commits Cause: Validation failed but rollback not executed Fix: Always rollback before retry
Symptom: Same validation error repeated indefinitely Cause: No max iterations or no error detection Fix: Enforce max_iterations=3, detect repeated failures
Input:
@skill exec
task: "create hello.py that prints hello world"
Execution:
Duration: ~20-30s Result: 1 commit with hello.py
Input:
@skill exec
task: "add authentication to Next.js app"
interactive: true
Execution:
Duration: ~5-8 minutes Result: 5 commits, working authentication
Input:
@skill exec
task: "add user profile feature: FastAPI endpoint + React component"
Execution:
Duration: ~8-12 minutes Result: 4 commits, working profile feature
v5.0.0 (2025-11-18):
Shannon CLI: Version 3.5.0 or higher (provides executor modules) MCPs: Serena (required), Sequential (recommended) Tools: Bash (for calling CLI commands), Read/Write (for file operations)
Installation:
# Ensure Shannon CLI installed
pip install shannon-cli>=3.5.0
# Verify exec modules available
shannon --version # Should show 3.5.0+
Status: Production-ready for Shannon Framework V5.0.0
This skill should be used when the user asks to "create a slash command", "add a command", "write a custom command", "define command arguments", "use command frontmatter", "organize commands", "create command with file references", "interactive command", "use AskUserQuestion in command", or needs guidance on slash command structure, YAML frontmatter fields, dynamic arguments, bash execution in commands, user interaction patterns, or command development best practices for Claude Code.
This skill should be used when the user asks to "create an agent", "add an agent", "write a subagent", "agent frontmatter", "when to use description", "agent examples", "agent tools", "agent colors", "autonomous agent", or needs guidance on agent structure, system prompts, triggering conditions, or agent development best practices for Claude Code plugins.
This skill should be used when the user asks to "create a hook", "add a PreToolUse/PostToolUse/Stop hook", "validate tool use", "implement prompt-based hooks", "use ${CLAUDE_PLUGIN_ROOT}", "set up event-driven automation", "block dangerous commands", or mentions hook events (PreToolUse, PostToolUse, Stop, SubagentStop, SessionStart, SessionEnd, UserPromptSubmit, PreCompact, Notification). Provides comprehensive guidance for creating and implementing Claude Code plugin hooks with focus on advanced prompt-based hooks API.